Police / Unions / Independent counting
How Many People Really Protested?
One procession, competing measurements—and a political struggle over the size of “the people.”
Would you judge the same protest differently if the headline used the other estimate?
410,000—or 800,000?
For October 24, the Interior Ministry reported 410,000 demonstrators nationally and the CGT reported 800,000. For Paris, the respective estimates were 110,000 and 250,000. These are attributed estimates, not a lower and upper statistical confidence bound.
Splitting the difference invents a new number without a measurement behind it. Show the provenance, scope and counting method before choosing a single headline.
BBC ↗24 October 2026 / attributed estimates
Read the data table
| Area | Interior Ministry | CGT |
|---|---|---|
| France | 410000 | 800000 |
| Paris | 110000 | 250000 |
A crossing count is not a photograph.
Occurrence uses elevated sensors defining a virtual line, supplemented by manual checks. Police count manually at strategic points and can review video. Unions also estimate rows and people per row.
A photograph measures a moment; a crossing count measures movement past a line over time. A pre-event forecast is neither. Joining downstream, repeat crossings, diversions and stopped groups can complicate interpretation.
AFP ↗Le Monde ↗Three approaches to counting
Police
Manual counting at strategic positions; video checking where available.
Organizers
Rows and estimated people per row; scope and method should be disclosed.
Occurrence
Virtual sensor line plus manual micro-counts and corrections.
Try the assumptions yourself.
Choose rows per minute, people per row and minutes observed. Their product is a modeled crossing count in a deliberately simple lab—not evidence about an actual event.
The exercise shows why assumptions compound. It does not claim that the actual police or union figures were produced with these settings, or that every crossing represents a unique participant.
Le Monde ↗How assumptions compound
10,800Modeled crossings—not verified unique attendees
15 × 12 × 60. Stops, repeat crossings and downstream arrivals complicate the result.
National is not Paris. Attendance is not violence.
For October 17, the national estimates were 550,000 according to the Interior Ministry and 800,000 according to organizers. On October 24, the official national estimate was 410,000 and the CGT estimate remained 800,000.
Compare the same geography and date. Do not add weekends to count unique individuals: some attended twice. Attendance estimates also tell us neither the number of looters nor the number injured or arrested.
BBC ↗The Journal ↗Compare the same national scope
Read the data table
| Date / France | Interior Ministry | Organizers / CGT |
|---|---|---|
| 17 Oct | 550000 | 800000 |
| 24 Oct | 410000 | 800000 |
Show the uncertainty.
Density, smoke, visibility and lateral movement affect counting systems. Independent sensors need manual corrections; independence is not immunity from measurement error.
An earlier expert review found the police method comparatively reliable and recommended reporting ranges. That review predated Occurrence’s media service; it cannot validate every later count or reject a method it did not study.
Le Monde ↗AFP ↗Where measurement becomes difficult
Ask for the observation window, counting location, corrections and scope of uncertainty.
The number is a political claim—and a measurement.
A high count can strengthen a claim to popular support; a low one can help officials describe dissent as limited. Those incentives do not prove any particular estimate is fabricated.
The most honest headline may contain two numbers. The most useful explanation gives their source, method, coverage and uncertainty—and says what neither number can establish.
Le Monde ↗Richard Martin ↗Before using a count
- 01Who issued it?
- 02Which date and area?
- 03Forecast, snapshot or crossings?
- 04What was corrected?
- 05What uncertainty remains?
- 06What does it NOT establish?
How to read this essay
Observed figures, dated reporting and legal texts are sourced. User-input calculators, city schematics and qualitative scenarios are labeled illustrations—not evidence about an individual or predictions. No synthetic data are presented as observed facts. Read the current legal text before relying on a provision.
Topics & Searches
- How Many People Really Protested?
- French politics visual essay
- French protests
- Police / Unions / Independent counting
- Amine Zidouh