Theme

Register of 2024 public money · GF0301

Police services

In short. Studies of the speed cameras deployed from 2003 find road deaths fell, from national series without a comparison group; a study of a city's cameras finds them marginal against crime.

2024

€28.9bn, of public money, line 19 of 155

Strength of evidence

Very limited level 1 of 4, on 1 of its 7 goals: Prevention and repression: reduce recorded crimes and victims · on rules or parts of the line, which have no separate cost

Studies

4 causal studies tied; 3 findings on its goals

Summary

Version 1 · written 19 October 2026 · history

What it is and who gets it

The national police and the gendarmerie, paid by the State, which spends most of this line, and the communes' municipal police.

Goals and the evidence on each

  • Prevention and repression: reduce recorded crimes and victims: Very limited, 1 finding
  • Strengthen activity against crime (clearance rates, trafficking, forensics, patrols, response time, training): no causal study found
  • Optimise use of mobile forces: no causal study found
  • Fight road insecurity: Very limited, 2 findings
  • Transparency and quality of the public security service (waiting times, satisfaction, ethics reports): no causal study found
  • Evaluate the tax expenditure (operational reserve, IJAT beneficiaries) (P176): no causal study found
  • Availability of gendarmerie air assets (P152): no causal study found

What the studies find

Of the 3 findings on its goals: 2 in the goal's direction, 0 mixed, 1 with no significant effect, 0 against it.

  • After the speed cameras of November 2003, the fatality rate per 100,000 vehicles fell by 20% (Laurent Carnis and Etienne Blais, 2013), and fatalities and injury crashes by 19.7% (Etienne Blais and Laurent Carnis, 2015); both studies follow national series with no comparison group.
  • In Montpellier, an evaluation of an extension of video surveillance, run in 2012-2015, found a marginal contribution against crime in public areas (Guillaume Gormand, 2017).
  • After a 2017 law let the national police use firearms beyond self-defence, its killings of vehicle occupants rose relative to the gendarmerie and the police of Germany and Belgium, though not significantly in stricter specifications (Sebastian Roché et al., 2025).

What is missing

  • No causal study was found of police staffing, neighbourhood policing (police de proximité, police de sécurité du quotidien), priority security zones or municipal police (search of 19 October 2026) (search notes).
  • No causal study measures the programmes' goals on clearance rates, mobile forces, waiting times or public satisfaction (the goals).

How this summary was made

Literature searched on 19 October 2026: causal evaluations of police and gendarmerie spending (staffing, neighbourhood policing, priority security zones, video surveillance and body-worn cameras, municipal police, speed cameras, rules on the use of force) since 2024. Searched: the HAL API (9 queries in French and English, 2024-2026); the OpenAlex API answered 8 queries with HTTP 429 (rate limit); the IPP site's search returned nothing on the police; interieur.gouv.fr refused automated access (403), and the full text of the Montpellier video-surveillance thesis is behind HAL's bot check. Strength of evidence rated by the rule on the Method page.

  • Version 1, 19 October 2026, written · by writer: First version. Latest updates

Under the register's two definitions

Broad definition: a causal study in the line

evaluated

At least one causal study is tied to this line, on all of it or on one of its rules; the line then counts in full, so this is an upper bound.

Strict definition: a single measure with its own cost and study

not yet evaluated

A spending category: the register holds no separate cost for the measures inside it, so none of its money counts under this definition.

4 causal studies are tied to this line, the latest from 2025. The share of public money under each definition is on the evaluations page.

Against the Haut-commissariat's perimeters of aid to companies:

  • not placed: the report lists no measure of this line

The perimeters, and the share of each with a causal study

The literature

The studies that touch this line

  • Laurent Carnis and Etienne Blais (2013). An assessment of the safety effects of the French speed camera program. Accident Analysis & Prevention.

    • Design, as recorded: interrupted time series
    • Goal measured: Fight road insecurity
    • Impact: effect « The fatality rate per 100,000 vehicles fell by 20% »
    • Leakage: not informed
    • Permanence: fades « a 26.2% reduction was recorded in the first month but dropped to 0.8 % for the last observation of the series »
  • Etienne Blais and Laurent Carnis (2015). Improving the safety effect of speed camera programs through innovations: Evidence from the French experience. Journal of Safety Research.

    • Design, as recorded: interrupted time series
    • Goal measured: Fight road insecurity
    • Impact: effect « the ASEP was associated with a decrease of 19.7% in traffic fatalities and crashes with injuries »
    • Leakage: not informed
    • Permanence: not informed
  • Guillaume Gormand (2017). L'évaluation des politiques publiques de sécurité : résultats et enseignements de l'étude d'un programme de vidéosurveillance de la Ville de Montpellier. PhD thesis (theses.hal.science tel-02439529).

    • Design, as recorded: impact evaluation with a protocol stated to respect Campbell Collaboration standards (comparison design not detailed in the abstract; flag: confirm counterfactual in full text)
    • Goal measured: Prevention and repression: reduce recorded crimes and victims
    • Impact: no significant effect « a marginal contribution of CCTV cameras against crime in public areas »
    • Leakage: not informed
    • Permanence: not informed
  • Sebastian Roché et al. (2025). And the law relaxed the rules. A quasi-experimental study of fatal police shootings in Europe. International Review of Law and Economics.

    • Design, as recorded: quasi-experiment: national police (treated) vs gendarmerie and German/Belgian police (control groups), monthly counts
    • Goal measured: an outcome that is not one of the line's goals; its reading has no colour and is not counted
    • Impact: adverse « The monthly number of killings has significantly increased for the national police »
    • Leakage: not informed
    • Permanence: not informed
    • another outcome, not counted

Each study is coded by this site from the phrase quoted beside each verdict: impact on the goal it measures, leakage to people the law does not target, and permanence where the aim is a lasting change. The rows are in criteria.csv and studies.csv.

Searches

When the literature was searched

  • 2026-10-04: every line, for causal evaluations of its measures and rules. Searched: the HAL, Crossref, OpenAlex, PubMed and IDEAS/RePEc catalogues of research and publishers' pages, line by line, and a general web search engine until its allowance ran out; some research departments' sites refused automated access, so a study they publish may have been missed (data/literature/gaps). Found: literature/studies.csv.
  • 2026-10-19: causal evaluations of police and gendarmerie spending (staffing, neighbourhood policing, priority security zones, video surveillance and body-worn cameras, municipal police, speed cameras, rules on the use of force) since 2024. Searched: the HAL API (9 queries in French and English, 2024-2026); the OpenAlex API answered 8 queries with HTTP 429 (rate limit); the IPP site's search returned nothing on the police; interieur.gouv.fr refused automated access (403), and the full text of the Montpellier video-surveillance thesis is behind HAL's bot check. Found: nothing new to code: SSMSI methodological papers on recorded crime, legal and sociological work on municipal police, body-worn cameras and police-gendarmerie zones, a thesis on crime and platform jobs (not police spending), and an essay on speed enforcement (Carnis 2025); the speed-camera studies were reread from their HAL abstracts: interrupted time series with no comparison group, scored 2.

Back to every line of the register

Sources: Eurostat, Ministère de l'Économie et des Finances, Direction de la Sécurité sociale, and the publishers of the studies, each named in its panel.

The page is static HTML and reads in full without JavaScript. The script adds the tooltips, traces a flow through a diagram, opens in place the evidence kept on separate pages, shows and hides the diagrams on a phone, recomputes the wealth calculator, places you on the page that asks where you stand, and keeps the theme you pick: light, dark or high contrast.

Questions, ideas or corrections: open an issue at github.com/samoht/budget/issues. Who makes this site, and the corrections made to it, are on the Method page.

Every data file, with its columns, licence and checksum: the Data page. Every chart, by question: the index of the charts. Every line of the register: the register's index.

The evidence summaries of the register as they are written: Latest updates (Atom feed).

Text and data: Licence Ouverte / Open Licence 2.0, except third-party data, which keep their publishers' terms. Code: ISC. See LICENSE.md.

Version unreleased-ef7a7ba2, built on 2026-11-12 from ef7a7ba2.