Build a national public-body register that preserves institutional history

The problem is not a total absence of data — it is fragmentation

OFFICIAL FACTFrance already publishes a detailed report on State operators. The law requires information on budget appropriations and earmarked taxes, employment, debt, own-source income, payroll, cash and property, and provides for publication in an open, reusable format. ^[1] It would therefore be wrong to claim that no inventory exists today.

THE REAL GAPThe everyday phrase public bodyA convenient umbrella term rather than one single French legal status. It can cover several kinds of public establishment, authority, grouping, operator and other publicly controlled entity. is wider than the budgetary category of State operators. Relevant lists are spread across budget documents, legal registers, sector datasets, independent authorities and public-establishment directories. Bodies can also be renamed, merged, transferred or removed from the operator perimeter, making institutional history hard to reconstruct.

PROPOSALMeasure 3.18 therefore proposes a consolidated, historical national register, backed by an open dataset and a clear legal duty to maintain it. The Official Journal can establish the legal framework and record major legal events; the operational register should be searchable, versioned and machine-readable rather than a static annual PDF.

Define the perimeter before publishing a headline count

The first layer can reproduce the State-operator perimeter. ^[3] A second can cover national public establishments outside that budget category. Additional layers may cover independent authorities, public-interest groupings, permanent consultative bodies, publicly controlled companies performing specific public tasks and other national entities that meet published criteria.

The register must avoid two extremes. A narrow definition simply duplicates the operator report. An excessively broad definition mixes very different entities, minor subsidiaries, subsidised associations and local structures into one meaningless total. The better approach is a multi-dimensional classification: legal form, level of government, public control, budget category, mission, financing model and group membership.

Build on what the State-operator report already provides

The existing statutory operator report is a strong foundation. It covers State funding and earmarked revenue, jobs, debt and off-balance-sheet commitments, plus multi-year execution data for own resources, payroll, cash and property. ^[1] A national register should import and cite those datasets rather than forcing operators to type the same information into a second system.

The 2026 State-budget overview also shows 401,389 FTEs financed wholly or partly in State operators. ^[2] That number demonstrates the size of the budgetary operator landscape but does not answer “How many public bodies exist?” The register’s role is precisely to prevent employment counts, legal-person counts, administrative-unit counts and policy counts from being conflated.

Use stable identifiers, not names alone

A name is not an identity. Bodies can be renamed without disappearing; several legal persons can merge; a mission can move to a successor. The register should rely on existing official identifiers wherever possible and add a lineage identifier where institutional history spans several legal persons. Searching an old name should lead the reader to the current successor and the legal event that changed it.

The minimum data model that makes the register useful

The first release does not need a hundred fields. It does need the fields that make audit possible: legal form, founding instrument, sponsoring ministry, short mission statement, budget category, identifier, workforce, budget, public funding, payroll, major property, critical information systems, subsidiaries, creation date and end event where relevant.

Every material value needs a date. “1,000 employees” without a year quickly becomes misleading. Each number should carry its reference year, source and ideally a link to the original document. Missing data should be shown as missing, never silently turned into zero to make the database look complete.

A documentation layer can then connect audit reports, performance plans and reports, contracts with the State, financial statements, appointment decisions and transformation legislation. The public-body page becomes a gateway to institutional memory rather than a flat directory entry.

Official Journal, public portal and API serve different purposes

The Official Journal — legal framework and institutional events

Legislation can create the register, define scope, assign responsibility for updates and specify publication duties. Many creations, mergers and abolitions already depend on published legal instruments; the register should link to those instruments rather than reproduce the Official Journal as text. ^[4]

The public portal — understanding

A citizen should be able to search an organisation or a ministry and immediately see legal status, mission, sponsor, budget, workforce and institutional history. Visual maps can show relationships, but the core should remain a readable, exportable record.

API and open files — verification

Journalists, researchers and administrations need stable machine-readable data. Each release should be timestamped so that the public landscape can be reconstructed as of a past date. That historical layer makes it possible to tell whether a body genuinely disappeared or merely changed name.

Why this measure does not create a large direct annual saving

COSTINGThe register is transparency infrastructure. Building it costs money: data normalisation, identifier reconciliation, software, quality control and historical maintenance. It may later reduce duplicated data collection and, more importantly, make structural overlaps easier to detect. Those later savings belong to the merger or abolition measures that use the evidence, not to the register itself.

The appropriate financial status is therefore “structural effect — democratic control”, not an invented annual saving. The value appears when the register prevents double counting, reveals a hidden institutional dependency or shows that a supposed abolition was in fact a merger.

How to stop the register becoming obsolete within months

Launching a portal is easy compared with maintaining it. Every entity needs a responsible data producer. Sponsoring ministries can certify parts of the record; the Budget Directorate can supply operator data ^[3]; official business and administrative registers supply identifiers; Légifrance supplies the legal basis. A central team should focus on conflicts, lineage and perimeter changes rather than retyping every field.

The service should publish its own data-quality indicators: percentage of bodies with valid identifiers, dated budget, dated workforce, founding instrument, sponsoring ministry, transformation history and date of last verification. A record not checked for three years should say so visibly. Trust grows when uncertainty is exposed rather than hidden.

Objections that improve the design

“Most of this information already exists.”

Exactly. The register is justified only if it reduces the cost of joining those sources and provides a common identity and history layer. If it becomes one more database requiring separate manual entry, it fails its own test.

“Public body is too vague to support one national number.”

That objection is correct if the site publishes one unexplained total. The solution is to publish counts by defined perimeter and let the user switch between State operators, national public establishments, independent authorities and other layers. Every headline count must carry its definition.

“Centralisation could expose security-sensitive information.”

The register should contain publishable administrative, legal and financial information. Network diagrams, vulnerabilities, personal data and operational security details do not belong in it. Transparency does not require publishing the attack surface of a public institution.