Agentic Economic Modeling
arXiv:2510.25743
The paper proposes Agentic Economic Modeling, a framework that uses large language models to generate synthetic choice data and then corrects it with a small human sample to improve econometric estimates of demand elasticities and treatment effects.
Abstract
We introduce Agentic Economic Modeling (AEM), a framework that aligns synthetic LLM choices with small-sample human evidence for econometric inference. AEM first generates task-conditioned synthetic choices via LLMs, then learns a bias-correction mapping from task features and raw LLM choices to human-aligned choices, upon which standard econometric estimators perform inference to recover demand elasticities and treatment effects. We validate AEM in two experiments. In a large scale conjoint study, using only 10% of the original data to fit the correction model lowers the error of the demand-parameter estimates, while uncorrected LLM choices increase the errors. In a regional field experiment, a mixture model calibrated on 10% of geographic regions estimates a treatment effect of -6510 bps on the hold-out regions, closely matching the full human experiment (-608 bps). These results demonstrate AEM's potential to improve RCT efficiency and represent a step toward LLM-based counterfactual generation.