Policy Comparisons and Causality in an Agent-Based Model

This paper demonstrates how an open-source agent-based model (ABM) serves as a mechanistic and counterfactual approach to causality. We aim to assess how interventions in social welfare and housing impact production and inequality across diverse cities. Our study verifies that the ABM adheres to the parallel trend hypothesis and fulfills essential criteria as a causal model. We employ ex post DAGs to identify key paths influencing policy responses, facilitating the construction of an effective panel regression model to recover the ABM’s ex ante causal mechanisms. Through this exercise, we not only demonstrate the ABM’s robustness but also provide targeted recommendations for the best policy instruments across different cities.