Resources
Last reviewed:
Coding agents can now write much of the code behind an empirical project. It’s tempting to “vibe code”, but I firmly believe human judgement will remain instrumental to framing research questions, choosing appropriate methods, and interpreting the results. Here, you’ll find practical tips for working with coding agents, useful datasets, and people worth following to learn more about using these tools in economics research.
Datasets
A curated collection of data sources for empirical research in economics and finance.
Monetary Policy
-
United States
U.S. Monetary Policy Event-Study Database
Measure interest-rate and asset-price responses around FOMC announcements, press conferences and minutes.
-
Advanced economies & emerging markets
High-Frequency Monetary Policy Shocks
Compare monetary-policy surprises across advanced economies and emerging markets using a common methodology.
-
Euro area
EA-MPD: Euro Area Monetary Policy Event-Study Database
Study how financial markets respond to ECB policy announcements and press conferences.
-
United Kingdom
UK-MPD: UK Monetary Policy Event-Study Database
Analyse UK monetary-policy surprises and market responses to Bank of England communication.
Asset Pricing
-
Equities
Explore equity return predictors with supporting data and code for replication.
-
Factors
Use factor returns and research portfolios as benchmarks in empirical asset-pricing models.
-
Bonds
Open Source Bond Asset Pricing
Replicate corporate-bond asset-pricing research with shared data and code.
-
Factors
Compare equity factors across international markets to assess how well return patterns generalise.
Macroeconomic & Market Data
-
India
Access RBI statistics for research on Indian monetary conditions, banking and the wider economy.
-
Economic history
Study long-run relationships between credit, financial crises and macroeconomic outcomes.
-
India · High-frequency data
An unofficial Python client for retrieving NSE market data; useful for exploratory downloads that require validation.
-
International
Compare credit, banking, exchange rates and property prices across economies.
-
India · Daily market data
NSE Historical Reports and Bhavcopies
Use official daily market reports and bhavcopies to examine Indian trading activity and check downloaded observations.
Pro Tips
A few habits that you should follow when using coding agents for empirical research.
-
Take time over the plan
Have the agent propose a plan before it writes any code, then question it the way a referee would. Which data will we use? What assumptions are we making? What are the constraints? Make sure these choices to be clear enough to question before they become part of the code.
-
Write down what you intend to test
Write down the main specification and robustness checks before looking at the results. There’s still room to explore, but keep track of what changed and why. When writing up, distinguish the analysis you planned from ideas that emerged along the way.
-
Make the agent prove its work
Before an agent refactors anything, have it write tests that pin down known results, such as a replicated table or a key coefficient, and read every diff yourself. The agent writes the code; you own the results.
-
Get a separate review
Ask a fresh session to review the code, providing the research question and relevant assumptions. Have it look for mistakes and explain its concerns. Then work through those concerns yourself; the reviewer can get things wrong too.
-
Replicate in a second language
For results that matter, have the agent rebuild the analysis in a second language, such as Stata alongside R, and match the estimates to six decimal places. If the answers differ, investigate. Matching answers are reassuring, though two implementations can still share the same mistaken assumption.
-
Maintain high reproducibility standards
Generate tables and figures from the analysis and keep track of where the reported numbers come from. Document the data, software, and steps needed to rerun the project. The AEA Data Editor’s reproducibility guidance is a useful reference when preparing the work for someone else to reproduce.
Who to Follow
-
Start here
Alessandro Spina
Claude Code for Academics Slides
A clear first tour: setting up, giving the agent a memory between sessions, the habits that prevent most errors and keeping your data safe.
-
Claes Bäckman
Claude Code in VS Code for Academic Economists Guide
A step-by-step setup for economists working in Stata, R, Python and LaTeX, with advice on keeping the agent’s context lean.
-
Paul Goldsmith-Pinkham
Claude Code for Applied Economists Video series
An eight-part series for Markus’ Academy at Princeton, from data analysis and web scraping to writing and collaboration.
-
Pedro H. C. Sant’Anna
My Claude Code Setup Workflow guide
A complete research workflow built on planning first, adversarial reviewer agents and replication-first coding.
-
Aniket Panjwani
Guide to Codex for Economists Guide
Written for OpenAI’s Codex, but the core loop carries over to Claude Code: plan the work, let the agent run, inspect the diff and keep only what improves the project.