On the Causal Interpretation of Agent-Based Economic Models: Epistemic Entitlements and Inferential Use

Abstract

Agent-Based Modeling (ABM) has become an established technique in economics as it incorporates heterogeneity, interaction, and adaptation into the construction of artificial economies (Chattoe, 1996, Dosi and Roventini, 2019). Its growing adoption reflects its capacity to explore phenomena that resist analytical closure and equilibrium-based representation. Yet the capacity of ABMs to support causal claims about economic targets remains deeply contested. On one side, critics have argued that highly idealized ABMs fail to deliver robust causal explanations, either because iterative model construction risks overfitting and post hoc rationalization, or because the resulting simulations lack the stability and transparency required for causal inference (Arnold, 2014, Frey and Šešelja, 2018). From this perspective, ABMs appear epistemically fragile: at best exploratory, at worst narrative devices with limited explanatory force. On the other side, recent contributions have questioned whether identification-centered standards of causality, such as the recovery of stable causal parameters, invariant structures, or difference-making relations, are appropriate at all for models characterized by endogenous expectations, learning, interaction, and nonlinearity (Henschen, 2018, 2025). Thus, the apparent causal deficit of ABMs may not reflect a failure of the models, but a mismatch between modeling practice and inherited causal frameworks. This paper underlines a deeper methodological tension between dominant conceptions of causality in economics and ABM practice. While identification-oriented approaches in econometrics tend to treat causality as a system-level property to be isolated through invariance, screening-off, or policy-invariant counterfactuals (Granger, 1980, Woodward, 2003, Pearl, 2010, Heckman, 2008, Heckman and Pinto, 2022), ABMs secure epistemic warrants through iterative construction and controlled misrepresentation, where the relevant causal channels are generated endogenously by the model’s own dynamics rather than imposed ex ante (Humphreys, 2004, Edmonds, 2010, Moneta and Russo, 2014, Guerini and Moneta, 2017). Within this workflow, causal relevance is not exhausted by explicit representational inclusion. For instance, an ABM may omit an explicit model of expectation formation from its conceptual framework, yet reproduce empirical regularities commonly associated with its presence. In such cases, heterogeneity, interaction, and model granularity can generate effects often attributed to optimizing expectations. Such results support conditional causal claims only relative to the constructed mechanism and the inferential control exercised within the model, rather than as discoveries of an isomorphic causal structure in the target system. As such, this work re-situates causal claims about ABMs within an analogical and purpose-relative modeling practice, where causal validity is assessed relative to scope, representational limits, and the conditions under which model-based counterfactuals are transferred to target systems. Building upon an existing analogical account of ABM (Weisberg, 2013, Boge, 2020, Rusconi et al., 2025), the paper treats models as epistemic instruments whose causal claims are mediated by explicit mappings between conceptual assumptions, algorithmic procedures, computational implementation, and target phenomena. On this view, causal claims are neither simply validated by empirical fit nor invalidated by the absence of unique identification; instead, they are licensed conditionally, depending on modeling purpose and on the ontic level at which the claim is directed, from micro-mechanisms to meso-regularities to macro-patterns (Manzo, 2022, Rusconi et al., 2025). Thus, the paper engages explicitly with econometric conceptions of causality grounded in counterfactual policy analysis and structural stability (Heckman, 2008, Heckman and Pinto, 2022, 2024), while contextualizing them within ABM specificity. The resulting framework clarifies what different classes of ABMs are epistemically entitled to claim, for which purposes, and under which assumptions such claims may be extended to economic target systems. More broadly, the paper contributes to interdisciplinary debates on causality, modeling, and explanation by showing how analogical reasoning can govern the responsible extension of model-based causal claims without committing to strong causal realism or collapsing into methodological skepticism.

References

Eckhart Arnold. What’s Wrong with Social Simulations? Monist, 97(3):359–377, 2014. doi: 10.5840/monist201497323.

Florian J. Boge. How to infer explanations from computer simulations. Studies in History and Philosophy of Science Part A, 82:25–33, August 2020. doi: 10.1016/j.shpsa.2019.12.003.

Edmund Chattoe. Why Are We Simulating Anyway? Some Answers from Economics. In Klaus G. Troitzsch, Ulrich Mueller, G. Nigel Gilbert, and Jim E. Doran, editors, Social Science Microsimulation, pages 78–104. Springer Berlin Heidelberg, Berlin, Heidelberg, 1996. doi: 10.1007/978-3-662-03261-94.

Giovanni Dosi and Andrea Roventini. More is different … and complex! the case for agent-based macroeconomics. Journal of Evolutionary Economics, 29(1):1–37, March 2019. doi: 10.1007/s00191-019-00609-y.

Bruce Edmonds. Bootstrapping Knowledge About Social Phenomena Using Simulation Models. Journal of Artificial Societies and Social Simulation, 13(1):8, 2010. doi: 10.18564/jasss.1523.

Daniel Frey and Dunja Šešelja. What is the epistemic function of highly idealized agent-based models of scientific inquiry? Philosophy of the Social Sciences, 48(4):407–433, July 2018. 1552-7441. doi: 10.1177/0048393118767085.

Clive W.J. Granger. Testing for causality. Journal of Economic Dynamics and Control, 2:329–352, January 1980. doi: 10.1016/0165-1889(80)90069-X.

Mattia Guerini and Alessio Moneta. A method for agent-based models validation. Journal of Economic Dynamics and Control, 82:125–141, September 2017. doi: 10.1016/j.jedc.2017.06.001.

James J. Heckman and Rodrigo Pinto. Econometric causality: The central role of thought experiments. Journal of Econometrics, 243(1-2):105719, July 2024. doi: 10.1016/j.jeconom.2024.105719.

James J. Heckman. Econometric Causality. International Statistical Review, 76(1):1–27, April 2008. doi: 10.1111/j.1751-5823.2007.00024.x.

James J. Heckman and Rodrigo Pinto. Causality and Econometrics. IZA Discussion Papers, No. 15081, 2022.

Tobias Henschen. What is macroeconomic causality? Journal of Economic Methodology, 25(1):1–20, January 2018. doi: 10.1080/1350178X.2017.1407435.

Tobias Henschen. Some problems of causal inference in agent-based macroeconomics. Economics and Philosophy, pages 1–23, March 2025. doi: 10.1017/S0266267125000021.

Paul Humphreys. Extending Ourselves: Computational Science, Empiricism, and Scientific Method. Oxford University Press, New York, 2004.

Gianluca Manzo. Agent-Based Models and Causal Inference. Wiley Series in Computational and Quantitative Social Sciences. John Wiley & Sons, Inc, Hoboken, NJ, 2022.

Alessio Moneta and Federica Russo. Causal models and evidential pluralism in econometrics. Journal of Economic Methodology, 21(1):54–76, January 2014. doi: 10.1080/1350178X.2014.886473.

Judea Pearl. The Foundations of Causal Inference. Sociological Methodology, 40(1):75–149, August 2010. doi: 10.1111/j.1467-9531.2010.01228.x.

Massimo Rusconi, Davide Secchi, and Raffaello Seri. The analogical roots of agent-based modeling in economics and social sciences: The case of innovation dynamics. Journal of Economic Methodology, pages 1–24, August 2025. doi: 10.1080/1350178X.2025.2547639.

Michael Weisberg. Simulation and Similarity. Oxford University Press, 2013.

James Woodward. Making Things Happen: A Theory of Causal Explanation. Oxford Studies in Philosophy of Science. Oxford University Press, Oxford, 2003.

Raffaello Seri

Professor of Econometrics

My research interests include statistics, numerical analysis, operations research, psychology, economics and management.