Congressional-Trade Alpha — An Event Study
Category: Machine Learning / Quant Research
Date: April 2026
About this project
An event study testing whether copying US congressional stock trades is an edge. 5,763 House disclosures, 2022–2025, with a proper market-model estimation window and SPY benchmark. The answer is no: sign-adjusted abnormal returns are roughly zero to slightly negative at every horizon a copy-trader could actually act on. The only signal with a pulse is committee relevance — members trading their own committee's sector — but the t-stats are weak and costs would eat it. A clean, well-powered null result.
Challenges
Parsing messy disclosure filings, building a leakage-safe event window, and resisting the urge to torture a non-result into a signal.
Learnings
Event-study methodology, and that proving a negative rigorously is its own kind of result.