We analysed 2,300+ STOCK Act disclosures. Here is what patterns genuinely predict 30-day returns — and which signals are noise.
Congressional trading disclosures get a lot of attention because they feel like an information edge — a member of a committee overseeing an industry trades a stock in that industry, and it's easy to assume they know something you don't. We wanted to test that assumption directly rather than take it on faith, so we pulled over 2,300 STOCK Act disclosures and measured what actually happened to the underlying stocks in the 30 days that followed.
What held up
The clearest, most consistent signal wasn't any single trade — it was clustering. When multiple members of the same committee bought the same stock or sector within a short window, the subsequent 30-day returns were meaningfully better than a random basket of comparable stocks. A single disclosure is weak evidence. Three or four disclosures in the same name, from members with relevant committee assignments, in the same two-week window, is a much stronger pattern.
Timing relative to committee hearings also mattered. Trades placed in the 10 trading days before a relevant hearing or markup session showed a modest but statistically noticeable outperformance versus trades placed with no such proximity.
What didn't hold up
Individual trades in isolation were close to noise. A single senator buying a single stock, with no clustering and no committee relevance, performed roughly in line with the broader market — not meaningfully better or worse. The size of the disclosed trade also wasn't predictive; large-dollar trades didn't outperform small ones. And disclosures filed near the STOCK Act's 45-day reporting deadline — meaning the trade itself happened well before the public actually saw it — showed no edge at all, since by the time the market could react, too much time had already passed.
How we'd actually use this
The honest takeaway is that congressional trading data is a low-frequency, low-confidence signal that only becomes useful when you filter hard: look for clustering, look for committee relevance, and largely ignore single, isolated disclosures. Used that way, it's a reasonable input alongside other signals — not a shortcut to alpha on its own.