Prompt Engineering for Economics

Modern LLMs can read filings, summarize literature, transform unstructured data, and stress-test arguments faster than any research assistant. Economics, Finance-Economics, and Mathematics-Economics work is being rebuilt around them.

Where this is showing up in Economics

  • AlphaSense, Hebbia, and Bloomberg's Document Search & Analysis let analysts query millions of filings, earnings calls, and research notes in seconds.
  • NBER working papers and top finance/accounting journals increasingly use LLMs to extract sentiment and content from 10-Ks, Fed minutes, and earnings transcripts — building on earlier work like FinBERT and Loughran-McDonald dictionaries.
  • Bloomberg GPT and the open-source FinGPT project are building finance-domain models; the Fed, BIS, and IMF have released working papers on LLMs for forecasting and policy analysis.
  • Research tools like Elicit and Consensus are reshaping literature review for empirical economics and policy work.

Projects you could build in this course

  • A tool that extracts and compares key metrics across hundreds of 10-K filings
  • A RAG assistant trained on Federal Reserve publications or IMF reports
  • A literature-review agent that synthesizes findings across econ papers on a topic
← Back to Thinking With Machines