Pension reform protests: How to count the number of demonstrators?

On January 31, Marseille police counted 40,000 protesters against pension reform; the CGT union counted 205,000. In Nantes, the police said there were 28,000 people and unions said there were 65,000. The science of counting crowds is subtle and even more complex during demonstrations, which are characterized by crowd movement and changes in density. But why are the numbers so different?

world Examine the methods used by the police, unions and independent companies hired by French media outlets.

Police Manual Statistics

After about ten people passed, the police manually counted by “tapping” the counter. The French interior ministry said the counts were carried out by “dedicated staff” who were usually positioned above crowds “a few hundred meters from the starting point” of the demonstrations.

According to the Interior Ministry, a second elevated counting point could be set up “midway along the route” during large demonstrations. In addition, police can videotape demonstrations and conduct a recount later after reviewing the footage.

union estimate

Unions are counted manually without technical assistance. Activists in the parade counted the ranks of demonstrators one by one. “They can estimate at a glance how many people are on each line and then multiply the two figures,” according to French public broadcaster Franceinfo.

An independent company using technology

Since 2017, independent counting of protests has been conducted by Occurrence.It provided data collected during the demonstrations to several media outlets, including world.

Occurrence relies on “Eurecam” technology, which is designed for “security at airports or fan zones,” says its president, Assaël Adary. Sensors placed along the demo route create a virtual line. Everyone who crosses that line is counted by the sensor.

The movement of people during a demonstration is more complex than in a fan zone at a sporting event. Differences in density or visibility, for example, could mean a margin of error of between 30% and 40%, Adary said. However, the margin of error can be corrected by “manual microcounting”, especially when the density is high, or when things like smoke grenades are disrupting visibility.

“When we recounted the votes, the demonstrators were added almost systematically,” Adari said. “We communicated to the media at the end of the presentation a kind of spreadsheet where we could see the numbers we observed in eight-second increments, and the numbers given by the sensors and the corrected numbers.”

Of the 40 demonstrations, the numbers recorded by Occurrence were 15% to 20% higher than those recorded by the counties and far below the numbers reported by the unions.

Which method is more reliable?

Is it possible to determine which method is the most reliable? In 2014, a committee attempted to answer this question. Dominique Schnapper, professor at the School of Advanced Studies in the Social Sciences (EHESS); Pierre Muller, inspector general of the national statistical office INSEE; and Daniel Gaxie, professor of political science, were commissioned by the Paris Police Department to study optimal counting methods.

Despite an “inevitable margin of error”, they concluded that the systems used by police were the least flawed. However, the study’s authors recommend that police provide a range rather than a specific number. Also, the 2015 study did not consider Occurrence’s methodology because the company initiated independent counting in 2017.