Data & tools
Papers are one output. These are the datasets, tools, and interactive appendices that came out of them — the parts other researchers actually pick up and use.
Fin-GloVe
General-purpose word vectors are trained on Wikipedia and news. They don't know that "restatement" is ominous, or that "headwinds" is a hedge. Fin-GloVe is trained on financial text instead, so the geometry reflects how the language is actually used in filings and disclosures.
Built for research use, with the corpus and vectors available for download.
Novelty in finance research
Which finance papers are genuinely new, and does novelty help or hurt publication? We measured novelty and conventionality across the corpus of finance research, then built an appendix where you can explore the results paper by paper rather than reading them off a table.
With Donohue, Drechsler, and Jiang. Review of Finance, 2023.
Dealscan–Worldscope crosswalk
Loan-level data and international firm fundamentals don't share an identifier, which quietly blocks a whole class of cross-border credit research. This crosswalk closes the gap.
Built for International Lending: The Role of Lender's Home Country, now maintained as a standalone WRDS dataset.
Form 10-Q itemization
A 10-Q is a legally structured document that arrives as an unstructured blob. Getting from one to the other reliably, across decades of inconsistent filings, is the unglamorous step that most textual finance research depends on.
With Zhang, Du, Sun, and Donohue. ACM CIKM, 2021.