Links

Kong and coauthors survey automated research through April 2026: papers for as little as $15, and where the systems still fabricate results or misjudge novelty.

NBER working paper 33363. Start here for the argument about automating the production of finance papers.

Lopez-Lira and Tang score news headlines with LLMs and test return predictability on headlines dated after the model's knowledge cutoff.

He, Lv, Manela and Wu train ChronoBERT and ChronoGPT on only the text available at each point in time, to remove look-ahead bias at the source.

Melissa Dell's overview of deep learning methods for economics research, with a companion site listed under courses.

Kelly and Xiu's survey of machine learning in asset pricing. NBER working paper 31502.

Zhu and Wang's multi-agent pipeline for empirical economics: data audit, hypotheses limited to what the dataset supports, estimation, drafting and automated review.

Stock Claude Code, Codex and Kimi Code write 117 papers alone. They read well, and none clears a top venue's bar once reviewers open the code.

Ning Li scores 912 AI-written economics papers against 41 from the AER and AEJ: Economic Policy. Ideas explain 71% of the gap, execution 29%.

Miyai and coauthors hand a coding agent a baseline paper. It finds the limitations, runs the experiments and writes the follow-up, and the authors list the risks.

Anton Korinek's catalog of research tasks that language models can take on. NBER working paper 30957.

Ludwig, Mullainathan and Rambachan on when LLM output is valid evidence: no training leakage for prediction, a validation sample for measurement.

Glasserman and Lin separate look-ahead bias from a distraction effect by removing company identifiers from headlines.

Xu and Yang replicate 384 political science studies and 3,523 models with an automated workflow. With a replication package, 92.1% reproduce fully or largely.

The Gentzkow, Kelly and Taddy survey. Written before LLMs and still the baseline vocabulary for text methods in economics.

Lu and coauthors on an end-to-end automated research loop. Not finance, but the reference point for the automation debate.

Cross-country banking, credit, debt securities and exchange rate statistics from the Bank for International Settlements.

Euro area monetary, financial and banking statistics from the European Central Bank.

Factor portfolio returns built on a common method across many countries.

Factor and portfolio return series, including international and developed-market sets.

Code and data for reproducing published cross-sectional return predictors. A ready benchmark for automated replication.

Free country-level development and financial indicators with wide coverage.

The platform most business schools use to reach CRSP, Compustat and other licensed datasets.

A BERT model for financial sentiment. Small, open and a common baseline to beat.

The Python client for querying WRDS with SQL. The piece that makes a data pull scriptable.

Companion resource to Deep Learning for Economists.

Companion site to Korinek's work on using generative AI across the research process.

Open textbook with reproducible code for empirical finance.

The standards AACSB-accredited business schools are assessed against.

Publisher of the Journal of Finance. Submission and disclosure policies are posted here.

Research and policy positions from the body behind the CFA charter.

Publisher rules on AI use by authors, reviewers and editors. Elsevier publishes the Journal of Financial Economics.

EFMD's institutional accreditation for business schools, the main counterpart to AACSB outside North America.

The EU securities markets regulator.

New quantitative finance preprints, updated on weekdays.

Research from the Bank for International Settlements, a steady source of cross-country work.

New working papers across economics and finance.