The short answer: insider buying has historically carried modest predictive value. Insider selling has carried almost none. Both statements need heavy qualification.
What the research found
The finding is old and reasonably durable. Studies going back to Jaffe (1974) and Seyhun (1986), through Lakonishok and Lee (2001) and later work, broadly agree on a few things:
- Portfolios formed on insider purchases have modestly outperformed on a risk-adjusted basis.
- The effect is stronger in smaller companies, where insiders have a larger informational edge over the market and analyst coverage is thin.
- Senior executives — CEOs and CFOs — carry more signal than directors or large outside holders.
- Insider sales show little to no predictive power in most studies.
Two caveats that matter more than the headline:
- These are averages across enormous samples. A tendency measured over thousands of trades says almost nothing about any one purchase.
- Published anomalies tend to decay. Once a pattern is widely known and cheap to trade, it usually weakens. Form 4 data is free, instant, and machine-readable, which is precisely the condition under which edges erode.
Why buying would carry signal at all
The asymmetry is structural, not mystical.
An executive is already massively over-exposed to their employer. Their salary, bonus, unvested equity, professional reputation and future earning power are all tied to the same company. Every principle of portfolio construction says they should be reducing that exposure, not adding to it.
So when an insider buys more stock on the open market, they are doing something that is bad for their personal diversification and expensive in cash. There are not many pleasant explanations for that other than believing the stock is cheap.
Selling has the opposite structure. It has a dozen mundane explanations — tax bills, a house, a divorce, a child's tuition, simple prudence — and the base rate is enormous, because most senior pay is stock.
Where the data misleads people
This is where most insider-trading coverage falls apart, and it is not subtle.
Grants are not purchases
A code A transaction is a stock grant. The company handed the executive shares as compensation. They did not choose to buy, and they paid nothing. Counting it as "insider buying" is counting payroll as conviction.
Option exercises are not purchases
A code M is an executive converting options — often granted years earlier — into shares. It appears as an acquisition. No purchase at the market price occurred. And it is very commonly paired with a same-day code S sale of exactly those shares, which is functionally a cash-out, not a bet.
On any given day, exercises and grants together outnumber genuine open-market purchases by a wide margin. Any site that lumps them together will show you far more "buying" than actually happened.
Planned sales are not decisions
A Rule 10b5-1 plan is a trading schedule adopted in advance. A sale executed under a plan adopted eighteen months ago tells you what the insider thought a year and a half ago, at best.
Since March 2023, filers tick a box on Form 4 declaring this. Before that, the only evidence is a footnote. Treating planned sales as fresh bearish signals is a straightforward error, and it is very common.
We flag every planned trade, including on pre-2023 filings →
10% owners are a different animal
A biotech crossover fund buying into a private placement files the same code P as a founder buying on the open market. The dollar figures can be enormous and the signal is completely different — one is a negotiated financing, the other is somebody paying the market price.
Check the price against where the stock actually traded that day.
What is worth more attention
Cluster buying. One insider buying can mean anything. Three or more different insiders at the same company buying on the open market within 30 days is much harder to explain away — they are not coordinating, they are each independently reaching the same conclusion.
Purchases as a share of net worth. A $50,000 purchase by a CEO earning $300,000 is a bigger statement than a $2M purchase by one earning $20M. The "% of holdings" change gets closer to this than the raw dollar figure.
Discretionary trades only. Strip out everything made under a trading plan and you are left with decisions somebody made recently.
The honest summary
Insider buying is one input. It is a genuinely informative one, more so than most public signals, and it is free. It is not a strategy, it does not work reliably at the level of the individual trade, and the effect has probably weakened as the data became easier to access.
The most useful thing you can do with it is not to follow insiders into trades but to notice when something is happening at a company you already understand — and then do the actual work.