Insider selling vs insider buying: why one signal is real and the other mostly isn't

Insiders buy for one reason and sell for a dozen. Here's why insider purchases carry information, why most insider selling is noise, and how to spot the sales that do matter.

2026-08-05 · 11 min read · By Insider Option

There is a well-worn line attributed to Peter Lynch: insiders might sell a stock for any number of reasons, but they buy for only one. It is the single most useful idea in insider-trading analysis, and it explains almost everything about how to read Form 4 filings.

The asymmetry, stated properly

Consider a CFO who already holds $8m of her employer's stock, earns most of her compensation in equity, and has essentially all of her human capital tied to the company's fortunes. She is, by any sensible measure of portfolio construction, catastrophically over-concentrated.

Now she buys $400,000 more on the open market. Every rational reason to diversify pointed the other way. She did it anyway. That is a costly signal in the economic sense — it is expensive to send and hard to fake.

Compare her selling $400,000. Plausible explanations, none of which involve any view on the stock:

• Diversification — she is over-concentrated and knows it.
• Tax liability from a vesting event.
• A house, a divorce, tuition, an illness.
• Shares withheld automatically to cover withholding (code F).
• An option approaching expiry that must be exercised or lost.
• A 10b5-1 plan she set up eight months ago and has not thought about since.
• A company policy requiring sales through a scheduled window.
• Charitable giving or estate planning.

Twelve reasons to sell, one reason to buy. The signal-to-noise ratio differs by an order of magnitude, and no amount of clever statistics fixes it.

What the research finds

The empirical literature is consistent with the asymmetry. Lakonishok and Lee (2001), examining decades of insider activity, found that predictive power came essentially entirely from purchases, and was concentrated in smaller companies. Jeng, Metrick and Zeckhauser (2003) constructed portfolios from insider transactions and found meaningful abnormal returns to purchases — roughly 11% annually in their sample — while sale-based portfolios showed no comparable effect.

Two qualifications that matter before you act on those numbers:

• The effect concentrates in small caps. A director buying a $400m company is a larger fraction of that company's information environment than a director buying a $2tn company. In megacaps the signal is close to undetectable.
• It has likely weakened. Post-Sarbanes-Oxley two-day reporting means the data is public far faster, screeners are ubiquitous, and published anomalies tend to decay. The 1975–1995 magnitudes should not be assumed to persist.

The filter that makes insider buying usable

Raw "insider buying" screens are near-useless because they include compensation mechanics. Strip it down to purchases that represent decisions:

• Transaction code P only. Exclude A (grants), M (option exercises), C (conversions), F (tax withholding), and G (gifts). This one filter removes most of the volume.
• Material relative to the insider's wealth, not in absolute dollars. A $150k buy by a director earning $250k a year in fees is a conviction trade. A $150k buy by a CEO with $80m in stock is a rounding error and sometimes a deliberate optics exercise.
• Meaningful change in holdings. Use the "shares owned following transaction" column — a purchase that increases holdings by 30% differs from one that increases them by 0.4%.
• Cluster buying. Three or more unaffiliated insiders buying within a few weeks is the most robust version of this signal in the research. One buyer can be wrong or performative; a cluster is harder to explain away.
• Role. CEOs and CFOs see the numbers first. A CFO buying is the highest-information single case.
• Buying into weakness. A purchase after a sharp drawdown says something different from a purchase during a melt-up.

The insider sales that do matter

"Mostly noise" is not "always noise". Specific patterns are worth attention:

Discretionary, off-plan sales

Check the 10b5-1 checkbox. A sale not under a plan was decided recently, by a person with current information. That is a different object from a scheduled sale.

Plan termination or amendment

Underrated. An insider who cancels a scheduled selling plan has actively decided to stop selling — a mildly bullish tell. One who adopts a new aggressive plan shortly after a run-up has decided the opposite. Post-2023 SEC amendments require disclosure of plan adoption and termination, so this is now observable.

Unusual proportion of holdings

Routine diversification trims a few percent. An officer selling 60% of their position in one transaction is not diversifying, they are exiting. Always compute the sale as a fraction of holdings before the trade.

Cluster selling by unaffiliated insiders

Same logic as cluster buying, inverted — and importantly, you must exclude common vesting dates. Five executives selling on the same day because RSUs vested that morning is a calendar event, not a consensus.

Selling ahead of news

Visible only in hindsight, but it is the pattern enforcement actions are built on. Worth flagging retrospectively when you are assessing management quality.

Aggregate insider ratios: handle with care

Market-wide "insider sell/buy ratio" indicators get quoted as sentiment gauges. Two problems make them close to meaningless as usually constructed.

First, they conflate dollars with decisions. Insiders receive equity as pay, so in aggregate they are structurally net sellers always — the baseline ratio is strongly above 1 in every regime, including at market bottoms. Second, most such ratios do not filter transaction codes, so vesting-related withholding inflates the "selling" side mechanically.

If you want to use aggregate insider data, count distinct insiders making code-P purchases rather than dollar ratios. That series behaves far more sensibly — it tends to spike during severe drawdowns, which is at least economically interpretable.

Summary

• Buying: one motive, real information content, strongest in small caps, strongest when clustered, strongest from CFOs.
• Selling: many motives, mostly uninformative, worth reading only when it is off-plan, disproportionate, clustered, or follows a plan cancellation.
• Always check the transaction code and the 10b5-1 box before concluding anything. Most bad insider analysis fails at exactly this step.

You can browse code-filtered insider activity on Insider Option's insider trading page, and the deeper treatment of returns is in do insider buys actually predict stock returns?

This post is for informational purposes only and is not investment advice. See our disclaimer.