Women insiders, 2023-2024: a small sample, a clear pattern

1. The headline number, in context

Across roughly 3,572 declared insider acquisitions in our 2023-2024 EU and US sample, 128 were filed by women. That is 3.6% of declarations, against 96.4% filed by men. The number is not noise. It is roughly what board composition statistics would predict, given that most personal buy declarations come from C-suite executives and controlling shareholders, both of which remain heavily male-skewed.

For context, women hold close to one third of board seats on the largest EU and US indices, but the population of declaring buyers leans toward executives with significant personal capital deployed in the company, not toward independent directors making token purchases. The share of declared buys is therefore systematically lower than the share of board seats, and that holds across our sample.

This is the first caveat to take seriously. The 3.6% figure is not a measure of how many women trade. It is a measure of how many women hold the kind of position that triggers a declared personal acquisition large enough to clear the filing threshold of the relevant regime.

2. Why the sub-representation is structural

The EU MAR and US Section 16 frameworks require declarations from a defined set of insiders: directors, designated senior managers, large shareholders, and people closely associated with them. The population of people who can file is structurally narrower than the broader corporate workforce, and within that population, gender composition skews male.

Recent index-level snapshots place women at around 33% of FTSE 100, CAC 40 and S&P 500 board seats, with executive director representation much lower, in the 8 to 12% range depending on the index. Personal buying activity is also conditioned by free capital, share ownership guidelines, and tenure. None of those mechanically excludes women, but all of them compound a base rate that starts well below parity.

A useful way to read the 3.6% figure is therefore: the personal-buying population is roughly half as gender-diverse as the broader board, and the broader board is itself well short of parity. None of this is news, but the filings reflect it cleanly.

3. The cluster pattern

The pattern that does survive close inspection is cluster behaviour. In our sample, 63% of women's declared acquisitions land within an active insider cluster on the same issuer, defined as two or more insiders buying inside a five-trading-day window. The equivalent figure for men is 46%.

The most plausible explanation is again compositional. Women in the sample are over-represented in independent director roles relative to executive roles. Independent directors tend to file alongside other directors during informational events such as a board refresh, a refinancing, or a coordinated buyback alignment. Executives, especially founder-CEOs, are far more likely to file alone.

That shifts the interpretation. A cluster is generally a stronger signal than a solo filing of equivalent size, because it is harder to dismiss as a personal portfolio rebalancing. The data we have does not show that women trade differently as a matter of style. It shows that the trades visible from women tend to land in contexts where cluster behaviour is already more likely.

4. Ticket sizes, and what they really measure

Average ticket size is roughly 138 k EUR for women versus 3.3 M EUR for men. The gap is large, and it is almost entirely a role gap.

The right way to read the median ticket is as a proxy for the seniority and ownership stake of the filer. CEOs and controlling shareholders make up a meaningful share of the male declarations, and a single founder buying 20 M EUR of their own stock dominates any local average. Independent directors, who account for the bulk of female declarations, tend to file smaller, more symbolic positions.

We have not tried to regress ticket size on role inside this article, but the rough decomposition is clear from a quick look at the raw data. Filings by independent directors of either gender cluster in a 50 k to 500 k EUR band. Filings by executives and large shareholders sit one or two orders of magnitude higher. Average ticket size by gender is therefore mostly a restatement of role composition.

5. Performance: a sample too small to talk down or up

This is where the temptation to over-claim is strongest, and where the sample is weakest.

In our 2023-2024 window, the 90-day batting average is 48% for men's declared acquisitions, and 41% for women's. At first glance that looks like a meaningful gap. With n=128 for the women's group, the Wilson 95% confidence interval on 41% is approximately 33% to 50%. That interval comfortably contains the men's point estimate of 48%. We have no statistical basis to claim that women, in the aggregate, are worse stock pickers, and a one-sided narrative in either direction would be misleading.

The 2026-and-after window is where this becomes interesting. As board gender composition continues to shift, and as more women move into executive roles where declarations are larger and more frequent, the sample of female filings will grow. With n in the several hundreds rather than 128, the confidence interval narrows, and the role-versus-gender decomposition becomes tractable.

For now, the honest summary is: small sample, no meaningful skill gap visible, and any future analysis needs to control for role before claiming one.

6. What this means for the Sigma signal

The Sigma scoring model does not include gender as a feature. That choice is deliberate and predates this analysis. Three reasons matter.

First, the legally interesting variance is already captured by role, cluster context, transaction value relative to existing stake, and the issuer's recent filing history. Gender, at the level of evidence we have, adds noise, not signal.

Second, introducing gender as a feature on a small sample would be exactly the kind of spurious factor that backtest hygiene is supposed to reject. With n=128 in one of the two groups, any apparent edge would not survive walk-forward validation, and chasing it would be a textbook overfit.

Third, the goal of the scoring model is to evaluate the trade in front of us, not the trader. Two filings with identical role, cluster, ticket and timing should receive the same score. Anything else is a different product.

7. What to watch from here

The data lever that actually moves over the next two years is board composition. EU and US targets for female representation on boards are still tightening, and a non-trivial share of new appointments are going to women, many of whom are independent directors with the kind of profile that produces declared buys.

If the trend continues, the female-filer share is likely to drift from 3.6% toward something closer to 5 to 7% by 2027, on the same definition. At that point, the sample becomes large enough to do real work on it. Specifically, three questions become answerable:

• Whether the cluster effect persists after controlling for role.
• Whether the median ticket gap narrows as more women reach executive seats.
• Whether the batting-average gap is real, reversed, or simply non-existent once role is held constant.

None of those answers exist today. The current sample lets us describe the shape of the data accurately and resist the urge to over-claim in either direction. That, on balance, is the most useful thing a research piece on this topic can do in 2026.