Auto-Econ (Sentiment Algorithim Database)

Auto-Econ: An Open-Source Text Algorithm Database for Monetary Policy Research

• Details of the public release coming by Fall of 2025

• Authors: Cory Baird*, Vincent P. Marohl†, Moritz Pfeifer‡, Benjamin Treitz§

• Research Focus: Addressing key challenges in applying Natural Language Processing (NLP) to monetary policy research
  • Upstream Problem: Transforming raw text into interpretable and structured numerical indicators
  • Downstream Problem: Ensuring valid statistical inference
• Project Contributions: Developed an open-source text algorithm database (auto-econ)
  • Enables exploration of various text analysis methods
  • Supports both traditional lexical models and recent Large-Language-Models (LLMs)
• Methodology: Comparative analysis of text indicators for U.S. Federal Reserve communications
  • Utilized basic Vector Autoregression (VAR) model, the most common empirical model in monetary policy analysis
• Key Findings:
  • Significant variations in model outputs when changing NLP methods
  • Provided best practices for NLP methods in monetary policy contexts
  • Developed guidelines for selecting appropriate NLP tools for different text types and economic analyses