Insider transactions

Definition

Corporate insiders — officers, directors, 10%+ holders — must file their trades in company stock (Form 4, within two business days), creating a public record of what the people with the best information DO with their own money. The documented asymmetry rules the read: insider BUYING is informative (documented positive drift, strongest in small caps and for clustered purchases), insider SELLING is mostly noise (diversification, taxes, compensation mechanics — sales happen for many reasons, purchases for approximately one).

How it works / structure

• The data (engine-executable): Form 4 filings on EDGAR (transaction date, type, size, resulting holdings); transaction coding matters — open-market purchases are the signal class; option exercises, 10b5-1 programmed sales, and gift/estate codes are mechanical (the platform filters by code before any scoring).
• The evidence: Lakonishok-Lee (2001) — insider purchases predicted returns (concentrated in small caps); selling carried little information; CLUSTERING (multiple insiders buying within a window) strengthens the documented effect; C-suite purchases outweigh director purchases.
• The mechanics that de-noise: 10b5-1 plans (pre- scheduled trades — disclosure-flagged since the 2023 SEC amendments, with cooling-off periods) separate programmed from discretionary; post-lockup sales (event-ipo-lockups) are expected supply, not signal; sales into buyback announcements (event-buybacks) are a documented incongruence flag (fa-earnings-quality adjacency).
• Legal frame (facts): insiders may trade legally on public information under the reporting regime; trading on material nonpublic information is illegal for everyone — the filings ARE the legal channel’s exhaust.

When it applies

Long-thesis corroboration (clustered open-market buying alongside an improving fundamental thesis); management- conviction reads on beaten-down names (the classic habitat of the documented effect); incongruence screens (insider sales + aggressive accruals + buyback announcements — fa-earnings-quality composite); post-lockup size reading.

Risk profile & failure modes

• Sell-signal overreach: shorting on insider sales ignores the documented noise floor — the asymmetry is the entire evidence base.
• Small-cap concentration: the effect’s documented habitat is small caps; mega-cap insider trades move nothing and mean little.
• Two-day lag + drift horizon: the signal is slow (months, not days) — it corroborates theses; it does not time entries.
• Planned-trade contamination: scoring 10b5-1 sales as discretionary conviction reverses the meaning; code filtering is mandatory, not optional.

Evidence & limits

Reporting rules are SEC-documented (including the 2023 10b5-1 amendments); Lakonishok-Lee (2001) and a consistent successor literature document the buy-side effect and the sell-side noise. Effect sizes are strongest pre-2000s samples; the platform quotes the direction with current- sample humility and replays any conditioned rule.

Falsifiable-thesis examples

Illustrations only, not signals:

• “The clustered-insider-buying cohort (3+ officers, open- market, 90 days) will outperform its size decile over the next year” — falsified by the cohort return.
• “X’s CFO purchase this month will be followed by no guidance cut for two quarters (information-content thesis)” — falsified by a cut.

Cross-references

• The frame: lens-sentiment (revealed conviction); institutional twin: sent-13f-holdings
• Incongruence composites: fa-earnings-quality, event-buybacks
• Supply-event separation: event-ipo-lockups
• The allocation lens: fa-capital-allocation (alignment)