Measure 1.14 — Publishing public-sector remuneration with useful, lawful transparency

Transparency already exists in limited form

Public pay and benefits are especially vulnerable to false costing: a national average says nothing about the distribution of high remuneration, and a tax value is not always an avoidable budget cost. Reform therefore needs post-level or homogeneous-group data with a clearly defined base and explicitly published exemptions.

French law already requires certain public employers to publish annually the aggregate of their ten highest remuneration payments with a gender breakdown. This provides a transparency baseline but is not equivalent to a comprehensive named database of all public-sector pay. ^[1]

An apparent reduction can also reappear as consultancy, outsourcing, another allowance, expense reimbursement or recruitment in an uncovered body. The perimeter must therefore be consolidated across administrations and operators. The meaningful result is full cost avoided at comparable service level, not the reduction of one payroll line.

Make pay visible without unnecessarily exposing people

A more detailed platform must respect data minimisation, accuracy, security and proportionality. Transparency can be granular on posts, grades, employers and pay components without unnecessarily publishing every employee’s identity. Publication thresholds must be justified by public interest. ^[2]

The legal vehicle shown in the historical corpus remains a reference point, but it is not copied without review. The chapter uses the level of law actually required: Constitution, organic law, ordinary statute, decree, chamber rule or management decision. This hierarchy matters because a misclassified reform can waste months in an unnecessary procedure or become legally fragile because an instrument of insufficient rank was chosen.

Build the platform from payroll systems

A robust model publishes pay scales, allowance components, senior posts above defined thresholds, the ten-highest-pay aggregates required by current law and statistical distributions by employer. An open API would enable analysis of disparities, while a correction mechanism handles errors. ^[1][2]

Implementation should be written before the rule takes effect: starting position, target rule, responsible authority, date, temporary exemptions, data to publish and control mechanism. The historical lead is Civil Service / Budget / CNIL, but the same actor should not produce all figures and certify its own result. A second line of review — Parliament, Court of Accounts, inspectorate, court or open data — must be able to reproduce the calculation.

Reproducibility is central to the Delta-Sierra approach. Readers should be able to move back from the conclusion to the starting data, understand assumptions and identify what still depends on a political choice. A reform can be ambitious without pretending all of its parameters are already known.

Cost comes before savings

This measure has no certifiable direct saving at launch: it initially creates costs for collection, standardisation, quality control, security and publication. Its return is indirect through anomaly detection, comparison of allowance policies and pressure for justification. Gains should be measured ex post, case by case. ^[3]

The calculation must distinguish a voted budget, executed expenditure and an assumption. An appropriation is not always fully spent; executed expenditure is not always avoidable; a behavioural estimate does not carry the same confidence as audited accounts. The final report should therefore assign a confidence level to each component and publish a range where the evidence does not support greater precision.

The Institutions block also contains strong interactions. Reducing parliamentarian numbers automatically affects some staff credits; abolishing a body may reduce support functions already counted elsewhere; capping pay can affect an allowance addressed by another measure. The financial ledger must assign every cash flow to one unique owner so double counting is impossible.

Deploy by major employer and control quality

The reference timetable has four stages. Stage 1 — zero baseline: freeze headcount, contracts, allowances, property, services, legal texts and indicators before reform. Stage 2 — rule and preparation: adopt the instrument, publish instructions and adapt systems. Stage 3 — transition: allow commitments to expire or transfer while respecting rights and service continuity. Stage 4 — stabilisation: measure a full twelve months on a constant perimeter.

Transition is not a footnote. It can include compensation, staff mobility, redistricting, IT adaptation, contract termination, property reassignment or training. Those costs belong in the dashboard because a reform with a high initial cost may still be worthwhile if recurring savings repay it; the reverse is also true.

Data protection and right to correction

Poorly calibrated transparency can expose individuals, facilitate harassment or spread inaccurate data. Conversely, aggregates that are too broad make the tool useless. The right balance combines thresholds, pseudonymisation or post-level publication, data controls and a rapid right to correction. ^[2]

A serious objection is used to design a safeguard. For Publishing public-sector remuneration with useful, lawful transparency, at least three indicator families should therefore be published: an institutional or service indicator, a budget indicator and a risk indicator. If savings rise while quality collapses, the measure is not validated. If quality improves but cost is merely shifted, it is not validated either.

A review clause should be built in from the start. Twelve to eighteen months after stabilisation, Parliament or the designated controller should compare target, actual outcome and unintended effects. Deviations are not a documentary failure: they are precisely the information needed to correct a reform rather than preserve a figure that has become unrealistic.

Build the dataset before deciding the saving

Costing should begin with a distribution, not an average. The reform needs to know how many people or posts sit in each remuneration or benefit band, which employers carry them, which components are statutory, variable or contractual, and which exemptions are legally justified. A sufficiently detailed aggregated dataset can support this calculation without publishing every employee’s personal information.

The simulator should then apply the target rule to that base and retain a trace of each correction: ceiling, exemption, transition period, job change, replacement cost or outsourcing. The sum of genuinely avoidable differences produces the savings range. This method is more demanding than a global percentage, but it can explain the result to a parliamentarian, union, financial magistrate or citizen using the same framework.

Dynamic effects must also be monitored. Pay rules change recruitment, mobility and contracting behaviour. After one and three years, the site should compare budget savings with vacancy rates, departures, external recruitment and purchased services. A saving that turns into shortages or expensive outsourcing should be corrected.

The public simulator that makes the costing falsifiable

A pay reform cannot be costed from a general average salary. It requires the actual distribution of relevant remuneration, separating base pay, bonuses, allowances, monetisable benefits and statutory situations, then applying the target rule individual by individual or by band. Where nominal data cannot lawfully be disclosed, an anonymised distribution by decile or range can still make the calculation auditable.

The public simulator should expose every parameter that changes the result: threshold, included pay base, legal exemptions, headcount, employer cost, implementation date and transitional effects. It should also show the counterfactual without reform. Readers can then verify that a saving comes from the proposed rule rather than a hidden assumption about departures, recruitment or bonuses.

Control must finally account for possible responses: moving compensation into an uncapped allowance, greater use of contractors, outsourcing, recruitment difficulty or relabelling a benefit. A gross saving is therefore never the final number. The relevant public number is the net gain after documented avoidance, adaptation costs and any targeted compensation requirements.

What must be demonstrated before retaining the structural effect target

The structural effect figure remains the Plan’s historical target here, not a budget receivable that has already been secured. To turn it into a bookable saving, the impact file should publish line by line: executed baseline expenditure, the share that can legally and operationally disappear, the year of disappearance, transition cost, expenditure taken over elsewhere and stabilised net saving. Where evidence is missing, the cell should remain open rather than being filled with an invisible assumption.

The proposed operational lead — Civil Service / Budget / CNIL — should provide that reconciliation in a reusable format with source documents and formulae. External reviewers must be able to reproduce the calculation without requesting additional confidential material, or identify precisely which protected data are indispensable. The final number may therefore be lower, higher or temporarily zero: Delta-Sierra’s first requirement is traceability of the result and the absence of double counting with other measures in the block.

← Back to the legal corpus of 155 measures