Eleven Sellers for Every Buyer: Why the Most Alarming Insider Number in 20 Years Means Almost Nothing

In the first half of 2026, executives at US companies sold 77.6 billion dollars of their own stock and bought 6.9 billion. Eleven sellers for every buyer, the second most lopsided reading in more than twenty years. In the same six months, a second insider statistic also hit a record – pointing in precisely the opposite direction. Both numbers are accurate. Only one of them contains information, and since 27 February 2023 you can read which is which directly off the filing instead of estimating it.

Insider data is one of the few inputs where a private investor receives exactly the same raw material as a professional: the same filing, at the same second, for free. That alone makes it worth learning to read properly. What follows takes the signal apart, identifies the portion that demonstrably works, and sets out a framework that lets you classify any single filing in a few minutes.

Two Records in Six Months

The first number comes from EPFR Global Market Intelligence and was reported in mid-July. Insiders sold 77.6 billion dollars of stock in the first half, a fifth more than the year before, exceeded only by the extraordinary year of 2021. Against that stood 6.9 billion dollars of purchases – barely above the 6.7 billion of the prior year, which at the time marked a seven-year low. Winston Chua, an analyst at EPFR, put it carefully: insider activity suggests executives are not especially eager to increase their exposure at current valuations. March alone produced 21 billion of sales against 2.3 billion of purchases.

The second number comes from SentimenTrader and was reported a week earlier. Across the constituents of the technology sector ETF XLK, 28 executives had bought their own company’s stock on the open market over the preceding six months. That is the highest count ever recorded; the previous record was 25, set in 2011. The figure had doubled since the start of 2026, in step with the retreat of many technology names from their May highs.

One data set, six months, two records pointing in opposite directions. This is not a contradiction in the data. It is a contradiction in the unit of measurement.

Measure	H1 2026	Context
Insider selling (value)	$77.6bn	up 20% on H1 2025, second highest in more than 20 years
Insider buying (value)	$6.9bn	marginally above the prior year’s seven-year low of $6.7bn
Ratio	roughly 11 : 1	measured in dollars, not in people
XLK executives buying	28	all-time record, prior high 25 in 2011, doubled since January

Why the Sell Side Is Structurally Blind

The asymmetry between buying and selling is one of the most durable findings in the empirical finance literature. Josef Lakonishok and Inmoo Lee examined every insider transaction on the NYSE, AMEX and Nasdaq from 1975 to 1995 and concluded that the informativeness comes entirely from purchases, while insider selling shows no predictive ability at all. In smaller companies, stocks bought by insiders delivered roughly 7.4% of abnormal return over the following twelve months, while the stocks they sold showed no meaningful underperformance. Jeng, Metrick and Zeckhauser confirmed the pattern in 2003 in the Review of Economics and Statistics using a different methodology: purchases earned abnormal returns of more than 6% a year, sales earned nothing significant.

The reason is mundane and sits in the compensation structure. An American executive receives the bulk of their wealth in stock and options of their own employer. Someone who sells is usually financing a house, settling a tax bill, funding a divorce, endowing a foundation, or simply reducing the concentration risk that arises when salary, pension and portfolio all depend on one ticker. There are a hundred reasons to sell, ninety-nine of which have nothing to do with the outlook for the business.

There is only one reason to buy. Someone who is already massively overweight, whose paycheck, retirement and net worth are tied to the same company, and who then commits additional personal money to that same stock in the open market, is acting against every rule of diversification. Nobody does that by accident.

Dollars Measure Wealth, Headcount Measures Conviction

That resolves the apparent conflict between the two records. The 77.6 billion dollars is a wealth statistic. It is dominated by a handful of founders and early investors whose holdings are so large that a programmed disposal of a few percent per quarter moves the aggregate. The 28 buying technology executives are a headcount. Each one counts exactly once, whether they bought 100,000 dollars or ten million.

An example from this year makes it concrete. At CoreWeave, one of the largest single contributors on the sell side, several executives sold during June and July 2026 – the chief strategy officer, the chief development officer, the chief executive and the chief financial officer. Every one of those transactions ran through a pre-arranged trading plan. In the sales executed on 24 June by two entities attributable to the chief strategy officer, the underlying plan had been adopted on 13 November 2025, more than seven months earlier. Whether the stock rose or fell that day had no bearing whatsoever on the execution. Those sales carry about as much information about CoreWeave’s future as a standing bank transfer carries about the account holder’s – and yet they land, at full value, inside the 77.6 billion dollar headline.

The practical rule that follows: when reading insider data in aggregate, count people, not dollars. And if you must count dollars, scale them to the individual’s existing holding and annual pay, never to the market capitalisation of the company.

What Changed on 27 February 2023

Until a few years ago, separating mechanical trading from deliberate trading required guesswork. That is exactly what Lauren Cohen, Christopher Malloy and Lukasz Pomorski did in a paper published in the Journal of Finance in 2012. They classified each insider by historical pattern: anyone who traded in the same month year after year was labelled a routine trader, everyone else opportunistic. More than half of all transactions fell into the routine bucket. A portfolio following only the opportunistic trades produced value-weighted abnormal returns of 82 basis points per month, while routine traders delivered essentially zero. Only the opportunistic trades predicted future firm-level news and events.

That was a statistical workaround. Since 27 February 2023 the answer is on the record. On that date the amendments to Rule 10b5-1 adopted by the SEC in December 2022 took effect. Four elements matter to an investor. First, a mandatory cooling-off period: for directors and officers, the later of 90 days after adoption or modification of a plan, or two business days after filing the Form 10-Q or 10-K for the quarter in which the plan was adopted, capped at 120 days. Second, a personal certification that the insider held no material non-public information when adopting the plan. Third, restrictions on overlapping plans and on single-trade plans. And fourth, the decisive one for whoever reads the filing: a checkbox on Form 4 indicating whether the reported transaction was made under such a plan.

Every filing now states whether it was pre-programmed. One caveat applies: the checkbox does not cover plans adopted before 27 February 2023.

The Study That Shows the Reform Worked – and What It Cost

Whether a rule works is not settled by its text. Sehwa Kim and Shivaram Rajgopal of Columbia Business School, together with Seil Kim of Baruch College, compared behaviour before and after the amendment. The most striking figure concerns the waiting period. Before the change, 31.1% of all plan sales occurred within 90 days of the plan being adopted – that was the actual loophole, because a manager who knew what the coming quarter looked like could adopt a plan and have it execute a few weeks later. After the change, that share was 1.7%.

Overall use of such plans fell only slightly, from 52.5% to 50.3% of insider sales. Among the group that had previously traded on very short timelines, usage dropped by 8.6 percentage points. The return finding is the one that matters: before the reform, plan sales were followed on average by significantly negative stock returns; afterwards, by flat or even slightly positive abnormal returns. Insiders also became significantly less likely to sell under a plan ahead of an earnings miss. The same reform closed a second pattern: because gifts must now be reported within two business days, the backdating of stock gifts that had previously been visible in price patterns disappeared.

The counter-argument sits in the same paper and deserves to be stated. For firms whose insiders had relied on short cooling-off windows, the informational efficiency of prices deteriorated, measured through intraday variance ratios. Regulators did not merely suppress abusive trading; they also removed information from prices. For an investor that cuts both ways: the remaining signal is cleaner, but it is also rarer.

Five Questions to Ask of Any Form 4

All of this collapses into a framework that runs in minutes. It does not replace analysis of the business, but it reliably discards what never needed to be looked at.

Question	What to look for	Why it matters
1. Buy or sell?	Only pursue purchases	All measurable predictive power sits on the buy side
2. Is the plan box ticked?	Ticked means programmed – discard	Readable directly since 27 Feb 2023 rather than estimated
3. Which transaction code?	Only code P counts	M, F, A and G are compensation or tax mechanics, not decisions
4. How large relative to the person?	Share of existing holding and of annual pay	One million dollars is everything or nothing depending on who signed
5. How many people, which roles?	Several insiders, different functions, tight window	Predictive content rises when multiple insiders buy simultaneously

Question three deserves elaboration, because it is where most beginners go wrong. Every line on a Form 4 carries a letter. P denotes an open-market purchase and is the only code that describes a voluntary decision made with the insider’s own money. S is a sale. M is the exercise of options or other derivatives, F is shares withheld to cover the tax bill on vesting, A is a grant from the company, G is a gift. Ignore the codes and you will systematically mistake payroll mechanics for conviction – reading an option exercise with an immediate sale as a purchase followed by a disposal, when in reality all that happened was that somebody got paid.

What a Real Cluster Looks Like

February 2026 produced a textbook case. At the data security vendor Varonis Systems, the stock had fallen a further 15% on 4 February after a quarter that beat expectations but came with cautious 2026 guidance – roughly 60% below its October 2025 peak and close to its 52-week low. Over the following days four insiders bought in the open market: chief executive and chairman Yakov Faitelson bought 26,725 shares at 22.41 dollars on 9 February for about 599,000 dollars, his first open-market purchase in twelve months; director Avrohom J. Kess bought 17,800 shares at 22.29 dollars on 6 February for about 397,000 dollars; another director bought 5,000 shares at 22.54 dollars; and the chief technology officer added just under 3,000 shares. Roughly 1.2 million dollars in total.

All five criteria were met: purchases rather than sales, open-market transactions rather than grants, no trading plan, several people across different functions, and a five-day window. A comparable pattern appeared on 13 May 2026 at FTI Consulting, where three senior executives together bought 14,400 shares for around 2.08 million dollars near the 52-week low, the largest portion of it by the chief executive.

What is not a cluster is the thing that looks identical at a glance: five filings from the same company on the same day that turn out to be grants under the compensation plan, or three sales from three different people that all carry the trading-plan checkbox and merely reflect the same quarterly cadence.

The Six-Month Rule That Does the Work for You

One feature of US securities law explains why an insider purchase carries more weight there than almost anywhere else. Under the short-swing profit rule of Section 16(b) of the Securities Exchange Act, officers, directors and holders of more than ten percent must disgorge to the company any profit realised from a purchase and a sale within any six-month period – regardless of whether they possessed inside information. Intent is irrelevant; only the calendar matters.

In practice, an insider who buys today cannot exit for six months without economic damage. The purchase is therefore not merely an expression of opinion, it is an enforced commitment. For the investor reading the filing, the statute involuntarily supplies a minimum holding period – and conveniently defines the horizon over which the signal should be judged. It also aligns neatly with the academic evidence, which measures abnormal returns over six to twelve months rather than days.

There is a second timing consideration that runs in the other direction. Most issuers impose blackout windows around earnings, so purchases cluster in the weeks immediately after results. A filing that arrives the moment a window opens is systematically less informative than one that lands in the middle of a quarter, because the first may simply reflect a backlog clearing rather than a fresh decision. In the European Union, that blackout is not a matter of company policy but of law, which makes the same caveat sharper for anyone reading filings from listed European issuers.

The Same Signal Is Getting Quieter in Europe

While the United States sharpened its insider signal in 2023, Europe is moving the other way – and it is doing so this year. The reporting threshold for managers’ transactions under Article 19 of the EU Market Abuse Regulation originally stood at 5,000 euros per calendar year. The EU Listing Act raised the default to 20,000 euros and gave national regulators the option of going as high as 50,000. Germany’s BaFin took that option in full by general decree on 4 December 2025: since 1 January 2026 the German threshold is 50,000 euros per calendar year. The stated purpose was to relieve issuers and executives of compliance cost; legal commentary suggests roughly a third of previous notifications will simply disappear.

For an investor holding European equities or ADRs, the consequence is unwelcome. It is precisely the smaller purchases that vanish – and by the five questions above, those are not the unimportant ones. A supervisory board member of a mid-cap adding 35,000 euros was reportable until December and is not reportable now. A chief executive of a large-cap disposing of a million-euro tranche remains fully visible. The reform shifts visibility exactly toward the transactions that carry the least information.

Feature	United States	Germany
Filing deadline	two business days (Form 4)	three business days, to issuer and regulator
De minimis threshold	none – every transaction is reportable	€50,000 per calendar year since 1 Jan 2026 (previously €20,000)
Trading plan identifiable?	yes, checkbox since 27 Feb 2023	no comparable flag
Profit disgorgement on a six-month round trip	yes	no
Blackout before reports	set by company policy	30 days, imposed by statute

The threshold increase is a national option rather than an EU-wide rule, so the cut-off can now differ from one member state to the next. Anyone comparing filings across European markets from 2026 onward may be comparing data sets recorded at different resolutions. Before drawing any conclusion from the absence of a filing, establish which threshold applies to the issuer in question.

What the Signal Cannot Do

Three limits are hard, and popular coverage routinely walks past all three. First, the aggregate buy-sell ratio is not a market-timing tool. It has been lopsided for years while equity markets rose; the 2011 record on the buy side did not mark the bottom of a cycle either, it occurred in the middle of one. Deriving an imminent decline from an eleven-to-one reading is using a cross-sectional statistic as a time-series forecast.

Second, the signal is strongest exactly where it is hardest to act on. The documented excess returns sit predominantly in smaller companies with thin analyst coverage – which is also where spreads are wide, liquidity is scarce and single-name risk is high. At a mega-cap followed by forty analysts, an insider purchase is a footnote.

Third, the horizon is long. The studies cited measure over six to twelve months. An insider purchase says nothing about the coming weeks, and the person making it generally knows nothing about the next rate decision, the next tariff proclamation or the next move in oil. They know something about their own company. That is a great deal – but it is also all.

A methodological caution belongs alongside these. Some of the findings quoted here come from sample periods that ended before the studies were published, and anomalies that attract attention tend to weaken. The 82 basis points per month from 2012 should be read as an order of magnitude, not as an expected return for 2026.

One further limitation applies specifically to US taxable investors. Acting on an insider cluster is, by construction, a position taken on a six-to-twelve-month view, which places any gain squarely in short-term territory taxed at ordinary income rates. The same trade executed inside a tax-advantaged account keeps the entire spread. If a strategy has an edge measured in single-digit percentage points, the account it sits in is not a detail.

What Follows From This

The most alarming insider number in twenty years is a wealth statistic. It says that a very small group of very rich people continues to reduce extremely concentrated positions, mostly on a calendar set months in advance. It says almost nothing about what those people think of their businesses.

The interesting number from the same six months is the smaller one: 28 people, a record since records began, in a sector that came under visible pressure between May and July. Whether that becomes a good investment decision is not decided by the headline but by five questions anyone can answer in a few minutes – and since February 2023, the most important of them no longer has to be estimated. It is read off a ticked box.

The working rule, then: treat sales as noise until proven otherwise, and purchases as a data point until proven otherwise. A cluster of unplanned open-market purchases by several insiders from different functions in a stock that has already fallen is one of the very few signals in public markets that is simultaneously free, immediate and empirically supported. It does not replace valuation work. It is simply a good reason to start doing some.