A friend sent me a screenshot of a Form 4 a while back. A director at a company he owns had sold somewhere north of a million dollars of stock, and the question attached was some version of should I be worried. My honest answer was that the filing told me almost nothing, and that if the same director had bought a million dollars of stock on the open market instead, I would have stopped what I was doing to read everything about that company. Same form, same dollar figure, completely different information content. That asymmetry is the single most useful thing to understand about insider filings, and most people watching them have it backwards, because a big sale feels scarier than a big buy feels encouraging.
The reasoning is old and it holds up under data. An executive can sell for dozens of reasons that have nothing to do with what they think the stock will do. Taxes come due on vested shares. A kid starts college. A divorce needs funding, a house needs buying, a portfolio that is ninety percent one ticker needs fixing. Most of it happens through 10b5-1 plans that were scheduled months in advance, precisely so the sales carry no discretionary information at all. When you scroll a raw Form 4 feed, the overwhelming majority of the sell volume is this kind of housekeeping, executed on autopilot by people who may be perfectly bullish on their own company.
An open-market buy has no equivalent set of innocent explanations. When an insider spends their own after-tax cash to add shares of the company that already pays their salary, already dominates their net worth, and already controls their career, there is exactly one motivation that makes the trade rational, which is that they think the stock is going up. Nobody buys more of their employer to diversify. Nobody buys to cover a tax bill. The purchase is a costly, concentrated expression of belief from the person with the best seat in the house.
What the research actually found
This is one of the few market anomalies where the academic work and the practitioner folklore agree. Studies going back decades, Lakonishok and Lee among the most cited, found that insider purchases have historically predicted abnormal returns over the following six to twelve months, while insider sales in aggregate predicted close to nothing. The effect showed up strongest in smaller companies, which makes sense, since that is where the gap between what insiders know and what the market knows is widest. The sell side of the dataset was so contaminated by tax and diversification selling that it carried almost no signal on its own.
Two practical consequences fall out of that. First, if you are going to spend attention on insider data, spend nearly all of it on the buys. Second, the sales are not worthless, but they only become informative when you filter them hard, which I will get to.
Token buys versus conviction buys
Not every purchase means much either. Executives know the market watches these filings, and a small buy after an ugly earnings call is often closer to public relations than to conviction. A CEO earning several million a year who picks up fifty thousand dollars of stock the week after the price halves is sending a press release, and it costs them roughly what a nice vacation costs. I treat those as noise.
The filters I actually use look like this:
- Size against compensation. Pull the proxy statement, find the annual cash comp, and compare. A buy worth a meaningful fraction of a year's salary, say a quarter of it or more, starts to look like conviction. A buy worth a week of salary does not.
- Size against existing holdings. A purchase that grows the insider's stake by ten percent or more says considerably more than one that adds a rounding error to an already large position.
- Transaction code P only. Form 4 tags open-market purchases with code P. Option exercises and grants show up under other codes, mostly M and A, and they reflect compensation mechanics rather than a decision to deploy personal cash. Ignore them.
- Cluster buying. Three insiders buying within a couple of weeks is historically one of the strongest configurations in the whole dataset. Independent people with the same private view, each risking their own money, is about as good as public-filing evidence gets.
- Who is buying. CFO purchases have historically punched above their weight, which fits intuition, since the CFO sees the cash and the forecasts before anyone else. Officer buys generally beat director buys, though a director suddenly writing a seven-figure check is still worth a look.
The selling that does carry information
Once you accept that most selling is noise, the interesting question becomes which sales escape the noise. A few patterns have earned my attention over the years.
Discretionary sales outside a 10b5-1 plan sit at the top. Filings disclose whether a transaction was made under a plan, and a large unplanned sale, particularly one that liquidates a big percentage of the insider's total position, is a genuinely different event from scheduled vesting sales. Percentage of holdings sold matters far more than the headline dollar amount. An executive selling five percent of their stake for eight figures is diversifying. An executive selling sixty percent of their stake at any dollar amount is telling you something about how they want their net worth positioned.
Cluster selling is the second one. One insider trimming is a Tuesday. Four officers all reducing within the same month, outside their normal vesting cadence, is a pattern, and it is worth checking what the company has said lately about guidance before assuming it is coincidence.
The third is timing games around the plans themselves. Adopting a new 10b5-1 plan and selling under it almost immediately used to be a reliable tell, enough of one that regulators eventually added mandatory cooling-off periods, roughly ninety days for officers and directors, between adopting a plan and trading under it. Plan modifications and cancellations deserve the same suspicion. Someone who cancels a scheduled selling plan may quietly be telling you they expect higher prices, and someone who rushes a new plan into place may be telling you the opposite.
And one more that took me a while to appreciate. Selling into weakness, meaning an insider dumping shares after the stock has already fallen hard, is more bearish than selling at highs. Selling at highs is what everyone does. Accepting a bad price to get out anyway suggests the insider thinks the current bad price is still better than what is coming.
How I actually use this
The workflow is short. Filter to code P open-market buys. Throw out anything small relative to the buyer's comp and existing stake. Flag clusters and CFO activity specifically. Then, and this is the part people skip, treat the surviving names as research triggers rather than trade signals. Insiders are notoriously early. They tend to buy on valuation and internal trajectory rather than timing, and it is common for a stock to keep falling for a quarter or two after heavy insider buying before the thesis plays out, if it plays out at all.
The failure mode worth remembering is sector-level trouble. Insiders know their own company, and they have no particular edge on macro. History has plenty of examples of bank executives buying their own collapsing stocks all the way down through a credit crisis, sincerely and expensively wrong. Conviction buying tells you management believes. It does not tell you management is right, so the buy list is where the work starts rather than where it ends.
These days when someone sends me a scary-looking sale, I ask two questions before anything else. Was it inside a plan, and what fraction of their stake did it represent. Most of the time those two answers end the conversation, and the time saved goes toward the much shorter list of people quietly buying.