Research  /  April 2026 · Non-finding

Why congressional trades don’t overlap with insider buys (and what we learned trying to test it)

Published 2026-04-17 · Test abandoned at the sample-size gate · ← All research

The idea we wanted to test

There’s a plausible-sounding conviction-stacking hypothesis: when a company insider files a Form 4 buy and a member of Congress has also bought that ticker in the prior 90 days, the combined signal should outperform insider buys at tickers with no congressional activity. Two independent informed actors putting their own money on the same name. We pre-registered a 12-month backtest to check it. A two-week sample-size sanity check killed the premise before we got to forward returns.

The numbers that stopped the test

In an April 16–29, 2025 Form 4 window with a 90-day House congressional lookback, the raw counts were:

Linear projection to 12 months: around 0–20 overlap events total. Our pre-registered gate required ≥ 100 to power a headline comparison. Gate fired STOP. We didn’t run the full pull.

The two lists don’t touch

Why they diverge

Form 4 P-transactions are mostly insiders buying their own employer. That universe is dominated by small- and mid-cap issuers and special situations — regional banks, biotechs, closed-end funds, recent IPOs, REITs. A CFO at a $400M regional bank doesn’t show up on anyone’s congressional trade radar.

Members of Congress trade through diversified brokerage accounts, often managed by advisors. Their disclosed purchases cluster in mega-caps and sector themes: big tech, big pharma, airlines, industrials. Those are companies where insiders rarely show up in the P-transaction stream — open-market CEO purchases at Apple or Amazon are unusual events, not weekly occurrences. The two pools barely touch.

What this means

Not adding to the scoring model. There’s no signal to extract here — not a weak one waiting for more data, not a statistically-invisible-but-economically-real edge. The mechanics of who files Form 4s and who trades through a Congressional brokerage account mean the candidate pool for the combined signal is essentially empty.

Next up: Test 3, repeat-buyer reliability — when the same insider buys the same ticker multiple times within a short window, does the signal strengthen with each repeat? That’s a hypothesis where the base rate is clearly large enough to test.