Simulating Macroeconomic Expectations in Survey Experiments with LLM-based Economic Agents

Researchers have developed a framework using LLM-based economic agents to simulate macroeconomic expectations in survey experiments, demonstrating that these AI agents can generate expectation distributions comparable to human survey data. The framework successfully captures human-like reasoning patterns when equipped with personal characteristics, prior beliefs, and external information, offering potential applications for economic modeling and expectation formation research.

This research addresses a significant gap in economic modeling by leveraging large language models to simulate human macroeconomic expectations in controlled survey environments. The framework's ability to replicate human expectation distributions has important implications for economic research and policy analysis, as understanding expectation formation is crucial for predicting economic behavior and policy effectiveness.

The work builds on growing interest in using AI systems as agents in behavioral and economic simulations. Previous approaches relied on simplified models or human surveys with limited sample sizes; this LLM-based approach offers scalability and consistency. The researchers' finding that prior beliefs drive distribution matching while personal and external information shape qualitative reasoning suggests a nuanced understanding of how expectations form—a principle applicable to financial markets and consumer behavior prediction.

For the financial and crypto sectors, this framework has implications for market sentiment analysis and volatility forecasting. Understanding how economic agents form expectations about inflation, growth, and monetary policy directly influences asset pricing models and risk assessment. The validated ability of LLMs to generate human-aligned expectations could enhance algorithmic trading systems and risk management tools that currently struggle with expectation formation modeling.

The research explicitly acknowledges boundaries and limitations rather than overselling capabilities, suggesting responsible development. Future work likely involves testing the framework against real market data, integrating heterogeneous agent beliefs, and exploring applications in central bank communication analysis or stress-testing economic scenarios. The field should monitor whether this approach becomes adopted in institutional finance for sentiment analysis or macroeconomic forecasting.

• →LLM-based agents can generate macroeconomic expectation distributions matching human survey data across multiple experimental designs.
• →Prior beliefs are critical for matching aggregate distributions while personal characteristics and external information drive human-like qualitative reasoning.
• →The framework offers scalability advantages over traditional surveys while maintaining human-aligned expectation patterns.
• →The research identifies clear boundaries and limitations, suggesting responsible development of LLM agents for economic modeling.
• →Applications extend to market sentiment analysis, policy testing, and understanding expectation formation in financial markets.