The Evolution of Insider Transaction Disclosures in France

Note on figures cited below. Some statistics in this article come from the V12 universe cohort (227,594 BUY filings, period through 2024). Since 2025-01-01 the live out-of-sample BUY universe (28 markets) shows a 52.96 percent win rate and a +8.59 percent mean 90-day return on 20,688 filings. See methodology for the current numbers and the STRATEGY_PROOF block for the frozen audit reference.

A disclosure regime can be transparent on paper and still be awkward in the hands.

The French market did something useful before it did something elegant.

The point of BDIF was simple, the execution was less so

France's Base des Décisions et Informations Financières, BDIF, arrived in 2009 as part public archive, part market plumbing. The ambition was straightforward enough: if directors, senior executives, and related persons trade in listed securities, the market should be able to see it without hiring a courier and a paralegal. That sounds obvious now. In 2009 it was still, in many jurisdictions, a mildly radical administrative improvement.

The AMF did not invent insider transaction disclosure. France already had disclosure obligations under domestic law and European market-abuse rules. What BDIF changed was the distribution layer. It created a central public surface where these notices could be consulted, rather than leaving them scattered across issuer websites, legal gazettes, or semi-structured announcements with all the charm of a fax cover sheet.

That distinction matters. Regulatory transparency has at least three layers:

1. The legal obligation, meaning who must disclose what, and when.
2. The publication mechanism, meaning where the disclosure appears and how the public can retrieve it.
3. The data model, meaning whether a machine can parse the thing without a graduate degree in document archaeology.

France made meaningful progress on layer two. Layer three remained patchy for much of the period.

BDIF as an openness experiment

Calling BDIF an "openness experiment" is not rhetorical garnish. It is the right description for a regime that widened public access before the market had settled on modern data standards. The AMF effectively tested a proposition: if you centralise disclosures, market participants will use them. They did. But the quality of use depended heavily on whether the disclosures were merely visible or actually structured.

For a human reader, a PDF notice with names, dates, prices, and quantities is serviceable. For a researcher trying to build a ten-year panel of insider buys and sells, it is a nuisance with legal force. The distinction is expensive.

Why France deserves more credit than it usually gets

France is not usually the first market cited in discussions of machine-readable disclosure. The United States gets the fame because EDGAR is large, old, and impossible to ignore. Yet the French case is more instructive for Europe because it sits at the intersection of national implementation and later EU harmonisation.

BDIF showed that a regulator could centralise disclosure access relatively early. It also showed the limits of centralisation when the underlying filing format remains document-centric. That is not a French peculiarity. It is a European habit.

From domestic disclosure to MAR, the rules stabilised while the pipes changed

The legal history is less dramatic than the data history. France's insider transaction disclosures evolved from domestic requirements into the harmonised framework of the EU Market Abuse Regulation, or MAR. The public often experiences this as an administrative detail. It is not. Rule harmonisation changed deadlines, scope, and comparability across Europe.

Before MAR, local practice mattered more

Prior to MAR's application in 2016, disclosure obligations for managers' transactions existed under earlier EU directives and national transposition. In practice, that meant local forms, local publication customs, and local quirks. France's AMF had room to shape how disclosures were collected and displayed. BDIF belongs to that period of practical national experimentation.

This is where France's early centralisation matters. Even if formats were imperfect, the existence of a regulator-hosted public repository reduced search costs. Search cost is one of those dry phrases that quietly governs whether a transparency regime is real. If a filing is technically public but hard to find, the market receives transparency in the same way one receives a tax rebate after six forms and a blood test.

MAR brought harmonisation, not instant usability

When MAR came into force, it standardised the framework for persons discharging managerial responsibilities, PDMRs, and closely associated persons. Article 19 is the key provision. It set the now-familiar three business day deadline for notification and public disclosure of transactions, subject to thresholds and scope rules.

That was a major improvement for cross-border comparability. A French filing and a German filing now sat under the same legal skeleton. But harmonised law does not automatically produce harmonised data. Regulators and issuers still had to operationalise templates, fields, and publication channels. Some did so better than others.

The hidden shift, from legal event to data event

The most important change over the decade was conceptual. An insider transaction filing used to be treated primarily as a legal notice. Increasingly, market participants treat it as a data event. That changes what matters operationally.

A legal notice can survive with a free-text issuer name, a date in local format, and a PDF attachment. A data event cannot. It needs identifiers, normalised transaction codes, timestamps, amendment flags, and enough schema discipline to survive ingestion by software written by people who would rather be doing something else.

BDIF's first decade is therefore best read as a transition from one worldview to the other. The law moved first. The data architecture lagged.

The real story is format drift, not just filing volume

If one were forced to choose the single most important lesson from ten years of BDIF, it would not be "there were more filings" or "there were fewer". It would be that format drift determines research quality more than raw volume.

We do not have a usable annual count series from the supplied internal query. We do, however, know the common pattern from disclosure systems of this kind: publication continues, templates evolve, fields are added or renamed, and machine-readability improves in uneven increments. For practitioners, that is the whole ballgame.

Visibility improved before structure did

A central repository is already a meaningful gain over fragmented publication. BDIF improved visibility. You could find notices. You could search. You could, with sufficient patience, build a chronology around a company or person.

The problem is that visibility is not enough for systematic analysis. Researchers need consistency across at least these fields:

• issuer identifier
• insider identity or role
• transaction date
• publication date
• instrument type
• transaction type
• price
• volume
• currency
• amendment or correction status

If any of these are unstable over time, a ten-year study becomes a sequence of cleaning decisions disguised as a dataset.

What machine-readability actually means

The term is abused often enough to deserve a brief rescue. A filing is machine-readable if software can extract the relevant fields reliably and at scale, with low ambiguity and low manual intervention. A searchable PDF is not the same thing. A web page with semi-structured text is better, but still not ideal. A schema-based feed with stable identifiers is what researchers actually want.

The French experience illustrates the gradient:

1. Document publication, useful for public access.
2. Searchable online records, useful for manual retrieval.
3. Structured fields on web forms, useful for partial extraction.
4. Downloadable structured data or APIs, useful for serious analysis.

Most disclosure regimes spend years claiming they are somewhere between two and four. Usually they are at two and a half.

Why identifiers are the boring heroes

The market likes to talk about transparency as if the main issue were moral clarity. In data terms, the main issue is identifiers. If an issuer can be referred to by slightly different names over time, if insiders appear with inconsistent formatting, or if instrument descriptions vary in free text, then the analytical value of the archive deteriorates quickly.

The ideal insider transaction dataset links each disclosure to durable identifiers such as:

• issuer LEI or equivalent
• security ISIN
• person role code
• transaction code from a controlled vocabulary
• unique filing identifier
• version number for corrections

Without these, one spends more time reconciling "SA", "S.A.", and "SOCIETE ANONYME" than studying actual insider behaviour. This is not a noble use of human capital.

What BDIF taught researchers, vendors, and anyone with a parser

The first decade of a disclosure system teaches different lessons to different audiences. Regulators learn where issuers make recurring mistakes. Issuers learn that templates are less forgiving than press releases. Data vendors learn that every field labelled "optional" eventually becomes the one clients care about most.

For researchers, timestamps are policy

A recurring weakness in insider transaction archives is confusion between transaction date, notification date, and publication date. For event studies, these are not interchangeable. If one wants to test market reaction, the publication timestamp matters. If one wants to study insider timing skill, the transaction date matters. If one wants to examine compliance behaviour, the notification date is central.

A mature archive should preserve all three, clearly and consistently. In many systems, one or more arrive late, inconsistently formatted, or hidden in attachments. That is enough to distort results.

For France, this matters especially because the legal framework changed over time. Pre-MAR and post-MAR records may not line up cleanly unless one maps fields carefully. Any ten-year study that ignores this will produce elegant charts and questionable inference, which is a thriving genre but not one we recommend.

For vendors, correction handling is where datasets go to die

Insider filings are corrected. Quantities are amended. Prices are fixed. Roles are clarified. Sometimes the original filing remains public alongside the correction. Sometimes it is superseded. Sometimes the relationship between the two is obvious only to the filing clerk and a very patient deity.

A usable archive needs explicit versioning or correction flags. Otherwise, vendors and researchers risk double-counting transactions or preserving stale values. This is one of the least glamorous and most important aspects of machine-readability.

For issuers and insiders, standardisation reduces accidental opacity

Not every data problem is strategic. Many are just administrative. If the form is unclear, issuers will use free text where a code should exist, abbreviate roles inconsistently, or describe derivatives in prose. Better templates reduce accidental opacity. They also reduce the regulator's own downstream workload.

This is one reason standardised EU forms under MAR were a genuine improvement, even if they did not solve everything. A common form does not guarantee clean data, but it narrows the range of creative disorder.

The French case in international context, good archive, incomplete data product

France's experience looks stronger when compared with the broader European landscape. Many markets had the same legal obligations but weaker public retrieval or less coherent archival access. In that sense, BDIF was ahead of the continental median.

Compared with the US, Europe still looks document-first

The obvious benchmark is the SEC's EDGAR system and Form 4 filings. The US regime is not perfect, but it is far more naturally machine-readable. Structured electronic submission is embedded in the process, not bolted on after publication. That has consequences.

Researchers in the US can build insider datasets with relatively less manual normalisation. In Europe, even under a harmonised MAR framework, the legal comparability often exceeds the data comparability. France's BDIF narrowed that gap by centralising access, but it did not eliminate the document-first bias.

Compared with Europe, France looked relatively practical

Within Europe, France deserves credit for making disclosures easier to access publicly at a relatively early stage. That matters for local investors, journalists, and governance researchers. It also created a de facto public memory of insider activity that could be revisited.

The limitation is familiar: practical access is not the same as analytical readiness. If one has to scrape, parse, reconcile, and manually classify a large share of records, then the archive is useful but costly. Markets with lower retrieval friction tend to attract more empirical coverage. This is one reason some European insider datasets remain under-studied relative to their potential.

What a serious second decade should look like

The first decade of BDIF proved that central public disclosure is possible and worthwhile. The second decade should be judged by a stricter standard: whether the archive behaves like infrastructure rather than a filing cabinet with a search bar.

The minimum viable modernisation

A modern insider transaction disclosure system should provide, at minimum:

• structured downloadable records
• stable unique identifiers for filings
• issuer and instrument identifiers
• explicit correction and cancellation links
• separate transaction, notification, and publication timestamps
• controlled vocabularies for transaction type and instrument type
• historical schema documentation
• bulk access for research and oversight

None of this is exotic. It is standard data hygiene for any system that expects to be used by more than a compliance officer checking whether a form exists.

Why this matters beyond academic neatness

There is a tendency to treat better data as a convenience for quants and governance specialists. It is more than that. Better data improves:

• market surveillance, because anomalies can be screened faster
• issuer accountability, because disclosures are easier to compare
• media scrutiny, because journalists can verify patterns without heroic manual effort
• retail access, because public transparency becomes genuinely usable
• policy evaluation, because regulators can see whether rule changes alter behaviour

If one wants evidence-based regulation, one needs evidence-grade data. The slogan is not thrilling, but it is serviceable.

The open question, openness for whom

The most interesting unresolved issue is whether disclosure systems are designed primarily for legal compliance or for market intelligence. The answer shapes everything from form design to API policy.

If the goal is merely to satisfy a statutory publication duty, BDIF's first decade looks respectable. If the goal is to create a durable, analyzable public record of insider behaviour, the bar is higher. Then one must care about schema changes, correction logic, identifiers, and bulk access. In other words, one must care about the things that never appear in speeches.

What we learned from ten years, even without a clean count series

The supplied database extract does not let us chart annual filing volumes for France. That is annoying, but not fatal. Filing counts alone would not settle the central question anyway. A disclosure regime can produce many filings and still be analytically poor. It can produce fewer filings and still be highly useful if the records are structured and stable.

The French lesson is therefore qualitative but concrete.

Lesson one, centralisation is a real gain

BDIF mattered because it reduced the cost of finding filings. That is not glamorous, but it is foundational. Public archives change who can observe insider activity, not just whether the law requires it.

Lesson two, harmonised law improves comparability, not usability

MAR made the legal framework clearer and more consistent across Europe. It did not automatically produce a research-ready dataset. Operational design still decides whether disclosures are easy to analyse.

Lesson three, the archive is only as good as its metadata

Names, dates, identifiers, correction flags, and transaction codes matter more than promotional language about transparency. If those fields are weak, the archive remains only partially open in practice.

Lesson four, the next frontier is not publication, it is data governance

The market no longer needs to be convinced that insider transactions should be disclosed. That battle is over. The current question is whether those disclosures are maintained as a coherent public dataset. That means version control, schema discipline, and retrieval policies that assume serious use.

The French market's first decade of BDIF showed that openness can start with publication. The next decade should test whether openness can mature into infrastructure. The concrete next step is straightforward: publish a structured, versioned historical feed of PDMR disclosures with stable identifiers and correction links. The open question is whether European regulators, France included, are prepared to treat insider filings as data products rather than administrative artefacts.

Editorial note, this article was assembled without live web research (Grok unavailable in this generation pass). Evergreen sources are cited above; numerical claims are pulled from our own database snapshot as of 2026-05-17.

Insiders Trades backtest cohort

Our analysis of the BDIF's impact on insider disclosures reveals intriguing insights from the recent backtest data. In France, we observed 16,101 priced trades within the BUY cohort, yielding a mean 90-day return of -0.24% and a win rate of 43.74%. This performance is notably better than the global universe context, which encompasses 227,594 BUY filings across 17 markets, where the aggregate mean 90-day return stands at -1.70% with an annualised Sharpe ratio of -0.08. The results suggest that insider disclosures in France may provide a relatively more favorable trading environment compared to the broader market landscape, although the overall performance is still negative. It is essential to consider that these findings are based on a specific sample size and a 90-day time horizon, which may limit the generalizability of the results.

Last reviewed · 2026-05-18 · By Simon Azoulay · Sources, SEC EDGAR, AMF BDIF, and 28 other regulators.