Opportunistic vs. Routine Insider Trading: How to Filter Form 4 Signal

Every day, hundreds of Form 4 filings land on SEC EDGAR. Most coverage of those filings treats every open-market purchase as a meaningful signal—a CEO buys $50,000 of stock and a headline appears. But the research tells a more complicated story. Opportunistic insider trading, the kind where an insider departs from their established pattern to make a discretionary bet, generates substantial abnormal returns. Routine insider trading, the predictable compensation-cycle variety, generates essentially none.

The framework for drawing this distinction was developed by Lauren Cohen, Christopher Malloy, and Lukasz Pomorski in research published through the National Bureau of Economic Research and subsequently covered by the Harvard Law School Forum on Corporate Governance. Their core insight: if you strip out the routine trades, the residual set of opportunistic buys contains almost all of the predictive power in the insider-trading universe. If you don’t strip them out, the signal you’re reading is diluted by roughly half before you’ve done any other filtering.

This post walks through what that distinction means in practice, what the return evidence shows, how the 2022 Rule 10b5-1 overhaul reshuffled where opportunistic activity now appears, and a concrete checklist you can apply to any Form 4 in the MarketPeel feed.

The Problem With Treating Every Form 4 the Same

Imagine two Form 4 filings arriving on the same morning. Both show a Code P open-market purchase. Both are from senior officers at mid-cap companies. On the surface they look identical. But one insider has bought in December every year for the last four years—it’s part of a predictable compensation-management cycle. The other hasn’t bought a single share in two years and is now putting in $200,000 at a price near a 52-week low.

These two transactions carry completely different amounts of information. The first insider is following a script; the second is making a discretionary bet. Lumping them together (as most raw Form 4 aggregators do) floods the signal with noise. Research synthesizing decades of insider trading studies confirms that the Cohen, Malloy, and Pomorski classification “substantially improved signal extraction,” with opportunistic purchases generating abnormal returns roughly four times larger than undifferentiated buy signals.

For retail investors using MarketPeel data, this matters immediately: before you apply any other filter (company size, cluster buying, sector), ask whether the trade is opportunistic at all. If it isn’t, the other filters are working on noise.

What Makes a Trade “Routine” vs. “Opportunistic”

The Cohen/Malloy/Pomorski classification is straightforward once you know it. A trade is classified as routine if the insider has placed a trade in the same calendar month for three or more consecutive prior years. The presumption is that a trade falling in the same month, year after year, reflects a pre-set schedule rather than a discretionary view of the stock. Any trade that departs from that pattern is classified as opportunistic—the insider chose to act outside their normal cycle, which suggests they had a reason to.

In practice, you can apply a simplified version of this test to any Form 4 without a database. Here are the indicators to check:

A Code P transaction with no 10b5-1 plan, in an unusual month, by an insider who last bought two years ago, is about as clean an opportunistic signal as you can find. A Code M (option exercise) followed immediately by a Code S (sale), both under a pre-existing plan, is as routine as it gets.

The Numbers: What the Research Actually Shows

The return evidence from the Cohen/Malloy/Pomorski study is striking enough to warrant quoting directly. Based on Thomson Reuters insider filings data from January 1986 through December 2007:

82 basis points per month compounds to roughly 10% per yearin abnormal return above a market benchmark—a figure consistent with the broadest academic literature on insider buying, but achievable only after the routine trades have been removed. The Harvard Law School Forum summary of the research notes two additional validation points: opportunistic traders are significantly more likely to face SEC enforcement action (confirming the classification is a genuine proxy for informed trading), and institutional investors mimicked opportunistic traders while providing liquidity to routine ones, meaning sophisticated market participants already apply this distinction.

Why Routine Trades Dominate the Filing Stream

Understanding why routine trades are so common helps you recognize them quickly. Several structural features of executive compensation push a large volume of transactions onto EDGAR that have nothing to do with a view on the stock:

RSU vesting on fixed schedules.Restricted stock units granted in a given year typically vest one, two, or three years later, often on a preset annual date. When RSUs vest, the insider receives shares automatically. This is a Code M transaction. Many executives then immediately sell a portion to cover taxes (Code F) or diversify (Code S). None of this reflects a view on where the stock is headed; it’s compensation management.

Annual gifting under Rule 144. Insiders who hold restricted securities frequently gift shares to family members or charitable foundations on a recurring schedule tied to tax planning calendars. These appear on Form 4 as Code G (gift) transactions and cluster predictably in December and January every year.

Pre-set 10b5-1 plans. An insider who adopted a 10b5-1 plan in January to sell a fixed number of shares each quarter will generate four Code S transactions per year on roughly the same schedule, indefinitely. If they do this for three or more years, those trades meet the definition of routine under the Cohen framework. Post-2023 research from Columbia Law found that even after the 2022 amendment tightened the rules, 10b5-1 plan usage declined only modestly from 52.5% to 50.3% of all insider sales, meaning pre-planned trades still constitute a majority of the sell-side filing volume.

Tax-withholding sales (Code F).When RSUs vest, companies typically withhold a portion of shares automatically to cover the insider’s tax liability. These Code F transactions appear on Form 4 but were never a decision by the insider; they are entirely automatic. They look like disposals in the raw data but signal nothing about the insider’s views.

Each of these categories floods EDGAR with transactions that are informative about compensation structure but not about the company’s future prospects. Learning to subtract them mentally before analyzing the residual is the first filter every Form 4 reader should apply.

What the 2022 Rule 10b5-1 Overhaul Changed

Before December 2022, the most common vehicle for opportunistic selling was the “single-trade plan”—an insider could adopt a 10b5-1 plan, designate a single large sale, and execute it within days. There was no minimum waiting period. This allowed insiders to effectively launder a discretionary sale through the safe-harbor structure.

The SEC’s 2022 amendments closed that specific loophole. Research published in 2025 on the Columbia Law Blue Sky Blog found the results were dramatic on the narrow technical point: the share of 10b5-1 sales occurring within 90 days of plan adoption fell from 31.1% to 1.7%. Sales that previously preceded negative abnormal returns shifted to flat or slightly positive abnormal returns, suggesting the most egregious informed selling had been removed.

But the signal didn’t disappear—it relocated. An analysis of 158,000 executive stock sales from 2016 through 2025 found that trades in the 90–120 day window after plan adoption rose from approximately 11% to nearly 30% of planned sales following the amendment. Annual abnormal gains in that window surged from roughly $23 million to nearly $89 million. Insiders adapted: instead of beating the cooling-off period, they land just beyond it.

Plan terminations remain the cleanest surviving signal. A separate Columbia Law study found that firms dropping their 10b5-1 plans after the amendment did so primarily because stricter rules blocked opportunistic trading, not because the compliance burden was excessive—the “opportunistic quitter hypothesis.” When an insider terminates a plan and then buys on the open market, that sequence carries meaningful information: they cancelled the pre-set plan in order to act discretionarily.

The Microcap Effect: Where Opportunistic Signal Is Strongest

The opportunistic/routine split matters at every market cap, but the signal magnitude is not uniform across firm sizes. A 2026 preprint studying 17,237 open-market insider purchases across 1,343 microcap issuers (market caps of $30M–$500M, from 2018 to 2024) found that the Form 4 signal is measurably stronger in smaller, less liquid companies. The highest-conviction subset in that study: purchases made after a stock had already appreciated more than 10% from its 52-week low generated a mean cumulative abnormal return of 6.3% with a 36.7% probability of outperformance over the subsequent period.

The dominant predictive feature in the gradient boosting model was distance from the 52-week high, accounting for 36% of total predictive signal. The implication is counterintuitive: an insider buying when the stock is already recovering from its low (not at the absolute bottom) is a stronger signal than one buying at maximum distress.

This aligns with what the original Cohen/Malloy/Pomorski work found about which insiders are most informative: the most predictive opportunistic traders tend to be local, non-executive insiders at geographically concentrated, smaller firms. In large-cap stocks, analyst coverage, institutional ownership, and information diffusion compress whatever edge an individual insider might have. In microcaps, that edge persists.

A Practical Checklist for Classifying What You See on MarketPeel

The following decision sequence applies to any Form 4 purchase in the MarketPeel feed. Working through it takes about 60 seconds per filing. Combined with the insider cluster buying filter(which checks whether multiple insiders are buying simultaneously), this approach produces the highest-conviction subset of signals in the filing stream.

• 1
  Is the transaction code P? If the code is M, F, A, or G, stop here. These are compensation-related and carry no signal. Only Code P (open-market purchase) is worth continuing to analyze.
• 2
  Is the 10b5-1 plan checkbox checked?If yes, check the footnotes for how long ago the plan was adopted. A plan adopted more than 90 days before execution is compliant with the 2022 rules. A sale landing in the 90–120-day window deserves extra scrutiny. A purchase under no plan at all is the cleanest opportunistic indicator.
• 3
  Does the insider trade in this same month every year?Pull up the insider’s prior filings on EDGAR or in the MarketPeel history view. Three or more consecutive years in the same calendar month = routine. Any departure from that pattern = opportunistic.
• 4
  How long since the insider last bought? A purchase after a multi-year gap carries more weight than one from an insider who buys every quarter. Gaps of 12 months or more between purchases are a rough signal of non-routine behavior.
• 5
  Is this a cluster?Once you’ve confirmed the trade looks opportunistic, check whether other insiders at the same company have bought in the past 14 days. An opportunistic cluster—multiple non-routine buys in a short window—is the highest-conviction combination in the filing stream.

Research on the cluster dimension finds that non-CEO insiders earn greater abnormal returns from opportunistic trades than previously recognized, extending the framework beyond the C-suite. A director or VP making an opportunistic buy deserves the same analytical attention as a CFO doing the same.

Limits of the Framework and What It Cannot Tell You

The opportunistic/routine classification is powerful, but it has real constraints worth stating plainly.

It requires prior trade history.The Cohen/Malloy/Pomorski method needs at least three years of historical filings to classify a trade as routine. For newly appointed insiders (a director who joined 18 months ago, or a CFO who took the role last year) you don’t have enough history to apply the test. You’re working with one or two data points. Treat these as ambiguous rather than automatically opportunistic.

Transaction costs compress returns at scale. A 2025 paper in Finance Research Letters using intraday data found that even genuine opportunistic signals compress significantly once position sizing and transaction costs are introduced at realistic scale. Returns from Form 4 signals are negatively correlated with stock liquidity, “almost negating a potentially profitable and scalable trading strategy even before transaction costs are included.” The signal exists, but it is strongest in less liquid stocks where it is also hardest to execute at size.

The signal and the legal line are close together.The original research found that opportunistic traders are significantly more likely to face SEC enforcement action—which is precisely why the classification works as a proxy for informed trading. An insider making a genuinely opportunistic buy may be acting on information that is material and non-public. The signal exists because some of these trades are well-timed; it should be read as an analytical input, not as an invitation to replicate. Research surveying 2025 literature, including a study on how CEO opportunistic trading encourages similar behavior among non-CEO insiders, suggests the opportunistic framework captures something real about information flow within organizations, not just scheduling noise.

Machine learning helps, but doesn’t solve the problem. A 2025 study highlighted in the Charles River Associates Q2 2025 literature watch found that ML models applied to Form 4 data boost predictive R-squared by 150% over traditional models and achieve 82% accuracy in predicting whether a trade carries informational content. That improvement is real. But 18% misclassification on a dataset where routine trades outnumber opportunistic ones means a meaningful false positive rate. The manual checklist above remains valuable even as automated tools improve.