Censorship of COVID-19 Demonstrations in France

by Antonios J. Bokas

Abstract: French protesters in anti-vaccine demonstrations from 2021 accused journalists of collaboration. However, an in-depth study of French internet news media revealed that journalists widely covered the events. Findings indicate that the internet continues to be a beachhead against oppressive censorship. Furthermore, strong support of unbiased journalism can reduce public distrust of the press and improve social discourse.

French internet news media widely publicized recent protests against COVID-19 restrictions. In early July 2021, protesters began demonstrating across France against President Macron’s COVID-19 health pass plan. The plan requires people to show a pass, which proves they are vaccinated or have natural immunity, to visit many establishments (Page, 2021). On July 18, a journalist from France24 called the protests a byproduct of “conspiracy theories” (Wheeldon, 2021, para. 1). Ten days later, Reporters Without Borders (2021) said one protester punched a journalist and another group of protesters harassed two journalists and called them “collaborators” (para. 5). In other nations, news media have marginalized those who oppose COVID-19 restrictions and objectors have also disparaged the press. However, a free press is the last defense against tyranny, and distrust of the press undermines civil discourse. Therefore, I sought to quantify coverage of the protests, detect any indicators of news censorship, and determine if claims of collaboration were true.

Framing the Issue

Accusations of disinformation and misinformation—tools used in information operations—have reached a high point since the COVID-19 pandemic began. However, pundits are probably unaware of the true meaning of information operations. In 2017, Facebook defined information operations as “actions taken by organized actors (governments or non-state actors) to distort domestic or foreign political sentiment, most frequently to achieve a strategic and/or geopolitical outcome” (Weedon et al.). Thus, French protesters ostensibly believed there was an organized effort to obscure opposition to COVID-19 restrictions in order to achieve political control or vaccine acceptance.

Their worries may be legitimate. France has already started to regulate “tech giants” that governments allegedly use to “crack down on dissent” (AFP, 2021a). For example, in July 2021, Autorité de la concurrence (Autorité, 2021), a French antitrust regulator, fined Google 500 million euros for removing content only from Google Actualités (the French Google News) to avoid paying royalties to French publishers (Ahmed, 2021). It was the second-largest antitrust fine ever issued to a single company in France (Ahmed, 2021). This seems to indicate the French government wants to defend the press. Therefore, the claim that they collaborated with news media to censor opposition to COVID-19 restrictions requires substantial proof.

The primary tool of the French government to enforce compliance with COVID-19 restrictions is a health pass called the pass sanitaire. A health pass, sometimes called a vaccine passport, is a digital certificate that proves an individual has been immunized against COVID-19 (Felter, 2021). France’s pass sanitaire also shows if an individual has tested negative for COVID-19 in the previous three days or recovered from it in the previous six months (Page, 2021). Although they strongly support COVID-19 vaccines, health security specialists Canyon and Kenavy (2021) concede that objectors have legitimate concerns about them, such as (a) past vaccine experimentation, (b) vaccine side effects, (c) the decline of vaccine efficacy over time, (d) the inability of vaccines to prevent transmission, and (e) their Emergency Use Authorization statuses. French objectors also believe that health passes are discriminatory and unduly restrict individual liberties (Page, 2021). In France, citizens need to display the pass sanitaire to eat at restaurants, patronize shops, visit hospitals, and board long-distance trains (Page, 2021).

With such a list of medical and legal concerns about a restriction that is socially and economically enforced, it is unsurprising that extreme opposition to the pass sanitaire quickly arose in France. Perceived censorship of opposition to such a system would also expectedly trigger civil retaliation. Thus, news censorship of protests against COVID-19 restrictions is an important matter for France and other nations that wish to peacefully implement orders or laws during medical emergencies. Pro- and anti-vaccine groups should research, vindicate, and defend the press to secure the tenuous threads that bind a democratic society together.

Research Methods

To improve the impartiality of my findings, I used both quantitative and qualitative research methods. For quantitative research, I employed web scraping to retroactively analyze news content. Web scraping is the use of a computer script “to download and process content from the web” (Sweigart, 2020, p. 267). I then tested the results for statistical significance using a two-tailed hypothesis test for two sample means. All web pages, scripts, logs, and worksheets collected or written during quantitative research can be found in the Open Science Framework repository at https://osf.io/wd6ht/?view_only=e6bd4dcffc774b18984d7eec4884dc58.

For qualitative research, I used structured analytic techniques as described by retired CIA analysts Heuer and Pherson (2015). First, I used Chronologies and Timelines to construct a general timeline of the French protests. The timeline was a central component of this study. After I completed quantitative analysis, I engaged in a Key Assumptions Check to reevaluate evidence based on statistical test results. Finally, I conducted Morphological Analysis to highlight additional scenarios in which censorship could occur. My intent was to advise future research about the causes of censorship. Internet archives greatly aided my assessment of this somewhat immeasurable phenomenon. My approach resulted in a formidable explanation of internet-based news coverage of the protests during the prescribed time period.

Timeline of Protests

The following timeline covers the period of July 7 to August 31, 2021. Statements that reflected broad sentiments at the time are paraphrased and presented in italics (see Figure 1).

Figure 1. Timeline of French Protests

  • July 7 – 13 President Macron announces his plan for a mandatory health pass (Bock, 2021). France fines Google €500M for violating rights of news publishers (Autorité, 2021).

  • July 14 – 20 18,000 protest the pass sanitaire across France (Bock, 2021; Thorburn, 2021). Protester: Health apartheid! (Douillard, 2021). Store owner: I’m not a bouncer (Gilles, 2021).

  • July 21 – 27 Protests increase: 161,000 march against the health pass nationwide (Bock, 2021). French Parliament approves a health pass bill (Howes & Allen, 2021).

  • July 28 – August 3 Up to 204,000 protest at 200 events; Organizer: We need a coup de force; Public officials: Protesters are lunatics (Bock, 2021; Feehan, 2021).

  • August 4 – 10 237,000 protest nationwide; some accuse the government of underreporting attendance (AFP, 2021b; Stanton, 2021). Collective: 415,000 attendees (AFP, 2021b).

  • August 11 – 17 Unprecedented protests continue as thousands attend over 200 events; Police: We expect 250,000 demonstrators (AFP, 2021b).

  • August 18 – 24 Protests continue with 1,400 participants in the Limousin and Corrèze regions; Protesters: Voluntary servitude, tyranny! (Jacquet & Rabiller, 2021).

  • August 25 – 31 Protests begin to decline; 70 people protest in the town of Chalon-sur-Saône and bring chairs, tables, food, and drinks to the event (Lambolez, 2021).

Some immediate conclusions can be drawn from the timeline:

  • Large protests began after President Macron announced his plan for a health pass.
  • Protest attendance peaked at the beginning of August after the pass was legalized.
  • Governmental and civilian sources reported drastically different attendance totals.
  • Protests were more peaceful in small towns and more violent in large cities like Paris (Douillard, 2021; Howes & Allen, 2021; and Bock, 2021).

Quantitative Findings

To quantify news coverage of the protests, I randomly selected 5 of the 8 most-visited news websites in France, Belgium, Germany, Italy, Spain, Switzerland, and the United Kingdom from SimilarWeb, a third-party website ranker, and analyzed them. Using an original Python script that accessed the Wayback Machine internet archive, I scraped two months of home pages from each website. I then compartmentalized each batch of scraped pages by nation.

With another original Python script, I parsed each batch of pages for the presence of keyword compilations. A keyword compilation is a set of words that must appear in a text element of the Hypertext Markup Language (HTML) of a page to produce a match. For example, if a keyword compilation consisted of “cat” and “dog” and a paragraph within an HTML document contained the sentence “I don’t want a cat or a dog,” then that paragraph would produce a match. However, if the sentence was “I don’t like your cat,” then it would not produce a match because it does not include the word “dog.” Based on a literature review, I determined which keywords would accurately represent headlines about protests against health passes. I used the following English keyword compilation to develop equivalent compilations for each language: “pass-,” “protest-,” and “demonstrat-” (see Table 1).

Table 1. Keyword Compilations

SampleBatchLanguageKeyword Compilation 1Keyword Compilation 2
2United KingdomEnglishpass-

I used stem forms of each keyword to yield diverse parse results. Each parse was also case-insensitive. For example, a query for the stem “pass-” (without a hyphen) would produce matches for the words “Pass” and “passport.” This approach enhanced parse objectivity and minimized the number of compilations needed to match language inflections. In addition, I selected keywords according to syntax and convention. Surprisingly, the core keywords remained quite homogeneous between languages. For example, the meaning of “pass-” is nearly identical in every sample language. Before I used the scripts for research, I extensively test them on various websites. The activity and results of each scrape and parse were written in .txt (text file) logs for future statistical analysis.

Statistical Model

I used a statistical model for two sample means to test my hypothesis and bolster the validity of the raw parse results. For a complete description of this method, see Chapter 8 of The Essentials of Statistics by sociologist Joseph Healey (2016).

Assumptions and Test Requirements

I used independent random sampling to construct each sample. Independent random samples are “gathered in such a way that the selection of the particular case for one sample has no effect on the probability that any other particular case will be selected for the other samples” (Healey, 2016, p. 228).

French news (sample 1) consisted of web pages scraped from the number 2, 3, 4, 5, and 8 most-visited news websites in France. Regional news­ (sample 2) consisted of web pages scraped from the number 1, 2, 3, 5, and 7 most-visited news websites in Belgium, Germany, Italy, Spain, Switzerland, and the United Kingdom. I set an initial target to scrape 2,170 pages, or 62 pages per website, from July 1, 2021 to August 31, 2021. Some pages were inaccessible due to script blockers or errors and dropped from the samples. However, the basic requirements for my statistical model were met: the sample size was large, a keyword compilation match was an interval-ratio level of measurement, and the sampling distribution was normal.


My model utilized standard formulas to test my research hypothesis against a null hypothesis. A null hypothesis is a statement of no difference between populations (Healey, 2016). The null hypothesis was  where  represented the French news population and  represented the regional news population. A research hypothesis is “a statement that contradicts the null hypothesis” (Healey, 2016, p. 444). My research hypothesis was  where the number of keyword compilation matches in  (French news) was postulated as being lower than that of  (regional news). I used the following formulas to conduct the significance test.

First, I calculated the standard deviation squared for each sample:

Then, I found the standard deviation of the sampling distribution:

Next, I obtained Z, the test statistic:

I finally compared the test statistic to Z(critical) to determine significance:

In this scenario, if Z(obtained) was lower than -1.65, French news coverage of protests against COVID-19 restrictions would be considered significantly lower than regional news coverage of similar protests with only a 5% chance of error. If Z(obtained) was higher than 1.65, French news coverage of protests against COVID-19 restrictions would be considered significantly higher than regional news coverage. If Z(obtained) did not fall in the critical region (below -1.65 or above 1.65), French news coverage of protests would not be considered significantly different from regional news coverage. I assessed that protesters believed French news media were part of an information operation that censored opposition to the pass sanitaire. This test would indicate if there was statistical merit to their fears.


Raw parse data and test results indicate that French internet news media extensively publicized the protests. I scraped and parsed 1,826 web pages from the Wayback Machine for an 89% script success rate. Sample 1 (French news) yielded 391 keyword compilation matches from 288 pages. Sample 2 (regional news) yielded 49 keyword compilation matches from 1,538 pages (see Figure 2). I calculated Z(obtained) as 9.77, which is well above the Z(critical) threshold of 1.65. French news coverage of protests against COVID-19 restrictions was significantly higher than regional news coverage of similar protests between July 1, 2021 and August 31, 2021 with only a 5% chance of error. I rejected the null hypothesis and my research hypothesis that French news covered the protests less than regional news.

Figure 2. Keyword Compilation Matches Versus Web Pages Parsed


The results of my hypothesis test only apply to the most popular news websites in the sample nations. Although the sample sizes were desirably large, my limited access to third-party website rankings reduced the sampling range. I initially wanted to sample 10 of the top 50 most-visited news websites for each nation. However, a lack of research funds blocked my access to these extensive lists. Thus, the statistical results may be loosely applied to websites outside of the population ranges but are not a conclusive representation of them. More funds would allow me to increase my sampling range and expand the application of my findings. For a description of sampling issues, see Healey (2016), p. 141. Despite these limitations, my findings are useful because they support qualitative analysis of news coverage in France.

Qualitative Findings

News content from July and August 2021 supports the statistics in Figure 2. France had the largest protests in Europe by far, followed by Italy and Spain (AFP & DPA, 2021; Reporters Without Borders, 2021; AFP & The Local, 2021). Although my findings prove one population of news media covered the protests, they do not describe the quality of their coverage. Of the French internet news articles I reviewed, I contend that (a) most fairly categorized the protests, (b) some were biased against anti-pass protesters, (c) few were biased against the government, and (d) few to none were biased against the pass sanitaire. Two salient questions remained:

  1. Why did some French protesters vehemently distrust journalists?
  2. Did other news media that I did not survey censor the protests?

To start answering these questions, I conducted a Key Assumptions Check with an objective “outsider” (see Figure 3) (Heuer & Pherson, 2015, p. 211).

Figure 3. Key Assumptions Check

Objector beliefs that news media collaborated with government to censor or disrupt protests likely stemmed from two sources: (a) the unethical practices of certain news outlets and (b) strong convictions that a health pass system violated basic human rights. Furthermore, hostile feedback between protesters and public figures almost certainly exacerbated tensions. In April 2021, President Macron said he would never use a health pass to divide the French people (Thorburn, 2021). Therefore, his announcement of a comprehensive health pass initiative in July, and Parliament’s approval of the pass less than two weeks later, likely intensified objector fears of censorship (Howes & Allen, 2021). Feehan (2021) reported that 65% of the French people supported a health pass. However, as Bock (2021) from The New York Times explained, “confused and erratic messages were commonplace” and hurt the government’s dialogue with objectors (para. 9). Thus, at a minimum, the French government approved the health pass without enough public debate. One consequence of the French government’s hasty implementation of the pass sanitaire is apparently 35% of the nation with an implacable distrust of institutions such as the press.


My findings present a number of positive developments for French news and government. First, they encourage the French people to have faith in their established news media and independent journalists. French citizens who oppose the pass sanitaire should actively share their views with internet journalists. Editors will probably publish their opinions in public fora. The number of times news articles quoted signage from protests was surprisingly high. Second, my findings isolated significant inconsistencies in the French government’s messaging about the health pass that contributed to skepticism. Most people will agree that any law that severely restricts routine social and economic activity will not be truly legitimate without profound public support. Therefore, I presented clear examples of what the French government did to aggravate public concerns and delegitimize its own plan. Lastly, my findings provide a map for future research into the field of news censorship.

As previously mentioned, my methodology for researching censorship could be greatly improved with additional resources. In fact, an effective future inquiry about censorship should simultaneously investigate all prevalent news media types on a given topic (refer to Figure 3, Assumption 5). I propose three possible scenarios wherein censorship could flourish (see Figure 4). A simultaneous investigation of each of these scenarios (and others) with methods similar to mine may reveal significant information.

Figure 4. Morphological Analysis: Three Possible Scenarios of Censorship by News Media

In Scenario 1, a corporately-owned television channel that earns ad-based revenue and has a large national audience may be prone to censor content that repels its viewers. Governments may also target such a news medium to influence its audience. This type of television channel would be more likely to practice censorship than a local, independent news channel that is funded by paid subscriptions or subsidized by viewer donations.

In Scenario 2, a government-funded news website that covers international topics may be likely to censor unwanted content. Such a website may be vulnerable to funding cuts if it does not comply with orders by corrupt public officials. Elections and catastrophes could sway politicians to enforce censorship by the website. A government-funded website at the local level would be less influenced by a political impulse to censor momentous, global developments.

Scenario 3 presents a dynamic situation in which either journalistic or corporate factors could impair or encourage censorship. If a local newspaper is funded by a community-oriented business, that business may encourage it to discuss controversial issues. However, the newspaper could be equally hamstrung if its sole benefactor was attacked by a competing firm. The benefactor could order the newspaper to censor harmful content to avoid humiliation and the newspaper would have little recourse. Journalism would take a back seat to favoritism.

This Morphological Analysis underscores the necessity of independent journalism. Sadly, it seems that some of the journalists that French protesters harassed and attacked were street reporters—the exact kind of reporters that may have listened to them. Censorship will probably never be a fully quantifiable phenomenon. By its nature, a search for censorship is a search for an absence of information. However, the methods I have described herein may help other researchers answer the pervasive 21st century question, “am I being censored?”


French internet news media widely covered protests against COVID-19 restrictions in July and August 2021. Quantifiable data collected from dozens of news websites and 1,826 web pages support this assessment. Although I detected some bias in French news reports, the majority of the articles I read were objective and fair. Many other nations will likely confront the divisive issue of COVID-19 health passes. Governments and citizens in those nations should learn from France and value, protect, and utilize internet news media to debate this issue. My findings indicate that the internet continues to be a beachhead against oppressive censorship. Future studies of censorship would benefit from my findings and from a wholistic approach that monitors news coverage of a topic on all viable media. Furthermore, certain news media may be more prone to practice censorship than others due to structural or organizational forces. Regardless, strong support of unbiased journalism would almost certainly reduce public distrust of the press and improve social discourse.


AFP & DPA. (2021, July 8). France, Italy see mass protests against COVID health pass. DW. Retrieved November 5, 2021, from https://www.dw.com/en/france-italy-see-mass-protests-against-covid-health-pass/a-58794976

AFP & The Local. (2021, August 7). Protests in Italy against Covid health pass. Retrieved November 9, 2021, from https://www.thelocal.it/20210807/in-pictures-italians-protest-against-new-covid-pass/

AFP. (2021a, September 18). Critics warn of Apple, Google ‘chokepoint’ repression. France24. Retrieved October 26, 2021, from https://www.france24.com/en/live-news/20210918-critics-warn-of-apple-google-chokepoint-repression

AFP. (2021b, August 14). Manifestations anti-pass: Cinquième week-end de mobilisation en France.

Ahmed, Z. (2021, July 13). Google fined almost $600M in France’s second-largest antitrust penalty. United Press International. Retrieved October 26, 2021, from https://www.upi.com/Top_News/World-News/2021/07/13/france-google-fined-news-content/3641626177982/

Autorité de la concurrence. (2021, July 13). Rémunération des droits voisins : l’Autorité sanctionne Google à hauteur de 500 millions d’euros pour le non-respect de plusieurs injonctions. Retrieved November 8, 2021, from https://www.autoritedelaconcurrence.fr/fr/communiques-de-presse/remuneration-des-droits-voisins-lautorite-sanctionne-google-hauteur-de-500

Bock, P. (2021, August 3). France brought in vaccine passports. People are not happy. The New York Times.

Canyon, D. & Kevany, S. (2021). Anti-vaccine sentiment – An existential disrupter of national and international security. Security Nexus: Daniel K. Inouye Asia-Pacific Center for Security Studies, 31.

Douillard, A. (2021, July 18). 150 contre le pass sanitaire. La Nouvelle République du Centre Ouest.

Feehan, K. (2021, July 31). Thousands of anti-vaxxers march through Paris holding placards saying ‘we are not guinea pigs’ and ‘it is our choice’ in opposition to compulsory jabs and vaccine passports. MailOnline.

Felter, C. (2021, April 7). Vaccine passports: What to know. Council on Foreign Relations.

Gilles, Q. (2021, July 14). Restaurants et bars digèrent mal le pass sanitaire. La Republique du Centre.

Healey, J. F. (2016). The essentials of statistics: A tool for social research (4th ed.). Boston, MA: Cengage Learning.

Heuer, R., Jr. and Pherson, R. (2015). Structured Analytic Techniques for Intelligence Analysis (2nd ed.). SAGE Publications; CQ Press.

Howes, S. & Allen, P. (2021, July 25). 100,000 march against France’s vaccine pass plan. Mail on Sunday (London).

Jacquet, F. & Rabiller, F. (2021, August 22). Sixième samedi de protestation contre le pass sanitaire. La Montagne.

Lambolez, R. (2021, September 2). 70 anti-pass sanitaire se rassemblent place de l’Hôtel-de-Ville; Chalon-sur-Saône. Le Journal de Saône et Loire.

Page, M. L. (2021). Do covid passports work? New Scientist, 251(3349), 12-13.

Reporters Without Borders. (2021, July 28). Reporters attacked while covering Covid pass protests in four EU countries. Retrieved November 5, 2021, from https://rsf.org/en/news/reporters-attacked-while-covering-covid-pass-protests-four-eu-countries

Stanton, A. (2021, August 7). Firefighters lead protests ahead of vaccine passports being required in France. Newsweek.

Sweigart, A. (2020). Automate the boring stuff with Python, 2nd ed. No Starch Press.

Thorburn, J. (2021, July 14). France’s Covid rebellion: Vehicles are set ablaze and tear gas is fired in Paris over Macron’s new law that all health workers get jabs and requiring new Covid pass for anyone over 12 going to a restaurant. MailOnline.

Weedon J., Nuland W., & Stamos A. (2017). Information operations and Facebook. Facebook. Retrieved on October 4, 2021, from https://i2.res.24o.it/pdf2010/Editrice/ILSOLE24ORE/ILSOLE24ORE/Online/_Oggetti_Embedded/Documenti/2017/04/28/facebook-and-information-operations-v1.pdf

Wheeldon, T. (2021, July 18). Conspiracy theories fuel French opposition to Covid-19 ‘health pass.’ France24. Retrieved on October 25, 2021, from https://www.france24.com/en/france/20210718-conspiracy-theories-fuel-french-opposition-to-covid-19-health-pass

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s