ConsumerSim

ConsumerSim: Generative social simulation reconstructs Consumer Confidence Index dynamics from synthetic populations

Researchers introduce ConsumerSim, a generative simulation framework that models Consumer Confidence Index (CCI) dynamics by constructing a microdata-calibrated synthetic population responding to macroeconomic, financial, policy, and news signals. The system outperforms persistence, time-series, and information-augmented baselines on U.S., EU27, and Japanese CCI reconstruction, with particular gains around high-salience economic shocks. Mechanism analyses reveal that CCI movements concentrate around salient events and that subgroup responses differ in magnitude across income, homeownership, education, and political-alignment dimensions. The work applies LLM-based generative agents to macroeconomic forecasting, contributing to the emerging literature on AI-driven social simulation.