On January 11, 2015, all over France, some 3.7 million people participated to rallies to protest against the terrorist attacks against the satirical magazine Charlie Hebdo and the Hyper Cacher supermarket. Their posters said “I am Charlie”, “I am Jewish”, “I am Muslim”, “I am a cop”, expressing solidarity with the murdered victims and attachment to freedom of expression (see Pelletier & Drozda-Senkowska, 2016). Few weeks later, however, other interpretations of the rallies took over. For instance, in a widely publicized and polemic essay titled “Qui est Charlie? Sociologie d’une crise religieuse” [Who is Charlie? Sociology of a religious crisis] (Todd, 2015), the demographer Emmanuel Todd, presented the rallies as an “imposture”, “a collective hysteria”, a “totalitarian flash”. According to him, the rallies did not represent the society at large, but mostly gathered an ageing, white, upper middle class and, above all, catholic France. In his analysis, the participants of the rallies are described as “catholic zombies”, who are not necessary churchgoers but who are impregnated by the authoritarian and hierarchical values of the faith they were brought up in. According to him, these participants actually did not defend freedom of expression against fundamentalism, they rather expressed the right to bash a minority religion, Islam (see also Zerhouni, Rougier, & Muller, 2016). Todd also argued that in a near future, this islamophobia could open the door to anti-Semitism and all other forms of racism.

Beyond the polemics between the “pro- Charlie” and the “anti Charlie”, the present paper proposes to examine the profile and motivations of the 11/1 marchers. Todd’s essay was based on the geography of the rallies, extrapolating from the social and religious structures and political traditions of the cities which had the largest numbers of protesters, with the risks of ecological fallacy it entails (Robinson, 1950)1. We argue, first, that survey data are more adapted to the exploration of individual characteristics than aggregated data; and second, that protests and street demonstrations cannot be fully understood without taking into account sociological and psychological theoretical existing models. Thus, in the present paper, the socio cultural, political and ideological profile of the participants to the Charlie rallies will be analyzed.

Our theoretical framework draws from the sociology of social movements and protest, from the pioneer work of Barnes and Kaase (1979) to the latest “CCC” project (“Caught in the Act of Protest – Contextualising Contestation”, Klandermans, Van Stekelenburg and Walgrave, 2014)2 comparing the profile and motivations of over 17 000 street demonstrators from 2009 to 2012 in 9 European countries. This literature shows that « elite challenging » modes of protest (demonstrations, boycotts, petitions), since the 1960s, have been mostly adopted by young (and now middle-aged), urban, educated, middle class citizens (Barnes & Kaase, 1979; Norris, 1999, 2002; Inglehart & Catterberg, 2002; Fillieule & Tartakowsky, 2012), in line with the development of “post materialist” values (Inglehart, 1977) based on autonomy and self-expression. These studies also indicate that every act of protest, every demonstration is unique. Each is called for by specific social movement entrepreneurs (the supply side) and takes place in a specific context that shapes its meaning (Klandermans, Van Stekelenburg, Walgrave & Verhults, 2012).

The general atmosphere of the post Charlie rallies was legitimized by all state authorities and institutions, and sponsored by the major parties and associations. Thus, it was seen very positively by mainstream media, and was associated with compassion, tolerance and solidarity in the public (Truc, 2016). Indeed, republican fraternity and freedom of expression were the main arguments put forward to encourage participation in the rallies (see Zaller, 1992; Kellstedt, 2003; Chong & Druckman, 2007 on the way public opinion is shaped by the way media and politicians frame the events).

However, Charlie Hebdo was known for its freedom of expression and irreverence on religious matters (Mignot & Goffette, 2015). Reputations are not necessarily based on facts. Indeed, analysing 523 front-pages of Charlie Hebdo from 2005 to 2015, Mignot and Goffette show that two thirds of them dealt with politics, only seven percent with religion, mainly Catholicism. Only 1.3 percent targeted Islam specifically. Nevertheless, because the journal reproduced one of the Danish daily Jyllands-Posten Muhammad cartoons, in 2006, and once again in 2011, it had the reputation of an islamophobic and blasphemous newspaper. Before the January 7 attack, it was brought to court several times by Catholic fundamentalist and Islamic associations and its journalists have been regularly threatened (see also Nugier & Guimond, 2016). This could make religious Muslims as well as Catholics quite ambivalent toward the rallies for Charlie, and place them at odds with the general atmosphere. This should even more be the case for French Muslims, who may not only be reluctant to take sides with a journal which caricatures the Prophet, but who may also fear to be pointed at and stigmatized for the attacks. Such situations of cognitive dissonance may hinder the expression of pro-Charlie sentiments in these two groups (Festinger, 1957; for an update Fointiat, Girandola & Gosling, 2013).

Based on the above considerations, we draw the following hypotheses:

H1: The Charlie marchers should show the usual socio demographic profile of demonstrators on post materialist issues, therefore be over represented among young, urban, educated, and middle class citizens (Van Aelst & Walgrave, 2001).

H2: The Charlie marchers should be overrepresented among left-wingers and among tolerant, non xenophobic citizens. Indeed, the post materialist values that drive protest and participation in demonstrations have been shown to be mostly values of tolerance and permissiveness, namely, values that are more frequent among left-wingers than among right-wingers (Quaranta, 2015). Moreover, as developed above, the very context of the Charlie demonstrations was framed in a tolerant, open way.

H3: Because the rallies supported a secular journal, for the reasons given above, Catholics as well as Muslims should be more reluctant to march for Charlie than non-believers.



To test these hypotheses, the ideal solution would have been to conduct a so-called INSURA” (“Individual Survey in Rallies”, see Fillieule & Blanchard, 2010). An alternative tool developed in France as soon as 1994 (Favre, Fillieule & Mayer, 1997), it consists in taking a sample survey among the participants, during the rallies. Since this could not be done, we used a standard opinion poll conducted two months after the attacks for the National Consultative Commission for Human Right (CNCDH) and the government, which included a question about participation in the rallies and demonstrations of January. A survey (conducted by the polling agency BVA) for this commission was conducted face to face, between March, 3 and 11, 2015, on a national sample of 1040 people. This sample was representative of the adult population living in metropolitan France, and was selected by the quota method (sex, age, occupation of head of household), stratified by region and category of agglomeration (Mayer, Michelat, Tiberj & Vitale, 2015).


Declared behavior can be very different from actual behavior (LaPiere, 1934; Wicker, 1969; Ajzen & Fishbein, 1980; Kraus, 1995; Fazio, 1990). Respondents may be tempted to give the expected answer, conform to the dominant social norms. The post Charlie demonstrations were so often and so consensually celebrated by the media, the elites and the political class, that it may have been difficult to go against the tide. Thus, some respondents may have lied about their actual participation. This is particularly plausible in a face-to-face survey as this one, where the interviewer is physically present, and the respondent may want to give the image of a “good” citizen (Kaminska & Foulsham, 2013).

In order to reduce these social desirability effects (Tourangeau, Rips & Razinski, 2000; Tourangeau & Yan, 2007; He et al, 2015), the word “Charlie” did not appear in the question wording, which just referred to demonstrations and rallies following the attacks. The question was: “Considering the demonstrations and rallies that took place in France after the attacks of last January, would you say that…”. Instead of a simple yes/no alternative, respondents could choose between three answers: “You took part in them”, “You did not take part but you would have wished to”, “You did not take part and you did not wish to”. The answers split the sample in roughly three thirds: 30 percent said they participated, 34 percent did not but wanted to, and 33 percent neither participated nor wanted to (3 percent did not answer). It is worth noting that a social desirability bias may have persisted. Indeed, if 30 percent of the respondents said they participated it would mean over ten million people marched on January 10–11, almost three times the estimated number given by the police and the media. However, socially and politically, the actual participants and those who wanted to be seen as such should not be very different. Conversely, one can be quite sure that people who declare their opposition to the “We are Charlie” movement did not so by chance: Such an answer being against the tide, we are confident their response should be accurate.


In line with what we know of the sociology of protest, Figures 1 and 2 document the contrasting profile of those who said they marched for Charlie, compared to those who did not approve the mobilization, those who wished they had participated standing somewhere in between.

Figure 1 

The sociological profile of the participants to the rallies.

Figure 2 

The political and ideological profile of the participants to the rallies.

The sociological profile of the participants to the rallies

As expected, the level of declared participation rises with the level of education and social status. Respondents with a university degree were four times more mobilized than those who only had a primary education. And the proportion of higher level executives and professionals who said they marched for Charlie is twice as high as the proportion of blue collar workers and more than three times than among small shopkeepers and craftsmen (Figure 1). This result is consistent with what was observed previously in most demonstrations on post materialist issues. Indeed, when blue collar workers protest, they usually march for their wages, their pensions or their social benefits, far less often for post-materialist issues such as antiracism for example (see Van Aelst & Walgrave, 2001; Fillieule and Tartakowsky, 2012). Even so, the overall level of declared participation of the manual working class is far from negligible (22 percent, i.e. roughly their proportion in the labour force). And if one combined the proportion of non-manual clerical workers and manual blue collars on the one hand with professionals and middle level executives on the other, the two groups were equivalent in size among the participants to the rallies. Contrary to the picture depicted by Todd, but as suggested by the sociology of protest, among the marchers, the younger cohorts born after 1976 were over represented: they mobilized three times more often than those born in 1940 or before.

As expected, there is a direct relation between the size of the city and the level of participation (Figure 1). Demonstrations are meant to attract many people and catch media attention; they usually take place in large cities and urban areas3. Whatever their social and ideological profile, respondents living in large cities will have more opportunities to participate to the rallies than those in distant rural areas or dilapidated suburbs (better information, easier transportation). More generally, dynamic urban areas attract young, skilled, connected and cosmopolitan populations, more likely to mobilize (Stoker & Jennings, 2015) than inhabitants of rural areas.

The political and ideological profile of the participants to the rallies

The political and ideological variables have the expected effect (Figure 2). Left wing respondents were far more inclined to take part in the Charlie rallies than the right-wingers. On the left, 43 percent said they participated, 31 percent that they would have wished to, and only 25 percent were against, whereas on the right, the respective proportions are 21 percent, 34 percent and 45 percent. This result is consistent with what was previously documented on the public of protests and more generally, on what used to be called “unconventional” political participation, which can nowadays be considered as “a normal way to act” among the young generations (Norris, 2002; Inglehart & Catterberg, 2002). Demonstrations, like other means of action such as petitions or boycotts, belong to the culture of “elite-challenging” participation. Such participations are more congruent with the culture of the Left than with that of the Right, which is more vertical and deferential to authorities. It is worth noting that probably because of the quite consensual dimension of these rallies, a rather large minority of rightist respondents seems to have participated in the rallies as well, joining forces with the usual routine type of demonstrators. Interestingly, the respondents who did not locate themselves on the political scale or chose the central position (named “Centre” in Figure 2) are not just in between those who did locate on the left or the right. Less opposed to the demonstrations than the latter, and more than the former, they also singularize themselves by their high level of “did not take part but wish I could have” answer (42 percent). This middle of the road position can be seen as an alternative for not choosing, as often the case for such middle modality of answer (like “neither agree nor disagree”). Their choice of answer could indicate that many of them are not really supportive of the Charlie Hebdo demonstrations, but more influenced by the social desirability norms, which prevents them to choose the third answer (“did not participate and did not wish to”).

The left-right contrasts are even more pronounced when party proximity is taken into account (Figure 2). Respondents who feel close to the radical left (“Left Front”, and Trotskyist parties) were by far the more supportive of the Charlie demonstrations, even more than the supporters of the mainstream left (“Socialist Party”, in power). At the other end of the spectrum, the radical right supporters (“National Front”) were clearly the most opposed: only 17 percent said they participated to the rallies, while 46 percent did not and did not wish too. The mainstream right supporters (“Union for a Popular Movement”, now labelled “The republicans”) were a little less reluctant than those of the radical right (22 percent took part, 40 percent did not and did not wish to), but clearly less supportive than their party leaders, especially their president Nicolas Sarkozy.

Further analyses were conducted in order to test xenophobia as a predictor of participation to the rallies. The xenophobia factor is the result of a principal component analysis (PCA) based on questions such as “migrants are a source of cultural enrichment” (agree/disagree), “there are too many immigrants in France” (agree/disagree), “we do not feel at home anymore in France as we used to” (agree/disagree). They all refer to migrants or strangers as a whole, not mentioning any particular religious or ethnic group. Respondents are sorted by quintile from the most xenophobic (++) to the most tolerant (– – –). The data contradict the idea that the real motivation of the participants to the Charlie rallies was Islam bashing and anti immigrant sentiment. There were prejudiced people among those who participated to the rallies, but as expected, the declared rates of participation were the highest among the most tolerant respondents. In the two first most prejudiced quintiles the rate of declared participation is 20 percent, but in the last and least prejudiced one it is 46 percent, with 28 percent saying they wished they could have taken part. The majority of the participants were not prejudiced, while the majority of those who did not participate were. It’s basically the same story if one looks at the level of islamophobia. Like for the xenophobia factor, the islamophobia factor is based on the results of a PCA of several questions regarding Muslims and Islam. Respondents were sorted by quintile from the most islamophobic (++) to the least islamophobic (– –). Those who had negative attitudes toward Muslims and Islam participated less than those who had no particular animosity towards this religion and its followers. The only intriguing result is the very slight drop of declared participation in the fifth quintile (from 37.5 to 36.5 percent), the most tolerant towards Islam. One would have expected these respondents to be the most mobilized of all. One of the explanations could be that in this quintile one finds all the respondents of Muslim faith, whose rate of participation, at least among the most religious, is lower than average (see infra).


Catholics represented 57 percent of the sample. Among them, regularly practicing Catholics are 10 percent (N = 57), occasional practicing are 21 percent (N = 125), the rest are non practicing. Respondents of Muslim faith represented 5 percent of the sample (N = 49). Among them, a little less than half are practicing Muslims.

As far as Catholics are concerned, in line with our third hypothesis, there is no over representation of them in the rallies, and specifically not of Catholic « zombies » (Todd, 2015), socialized to Christian values but not going to church anymore. On the contrary, the participation of Catholics as a whole was below average. Among them, the most mobilized were the occasional churchgoers. As can be seen in Figure 1, most of the participants were respondents with no religious affiliation. As far as Muslims are concerned, two main indicators were used. The first is the migrant origin of the respondent. Those born, or with parents and/or grandparents born, in Maghreb or Sub Saharan Africa, in majority Muslim countries, declared more often taking part in the Charlie rallies than those of only French or European ascent. The second indicator was declared religion. Although the number of declared Muslims is small in the sample and thus, one should be cautious, on the whole, Muslims declared participation was slightly above the sample’s average (by less than 3 percentage points) (Figure 1). However if the group is split between those who said they go to the mosque and those who do not, a clear cut cleavage appears. Indeed, 26 percent of the mosque-goers Muslims but 39 percent of the privatized Muslins (Brouard & Tiberj, 2011) went to the rallies.

A multivariate model

Cross tabulations sketch out a rough picture of the pro and anti-Charlie marchers. However, in order to validate our hypotheses, more sophisticated statistical analyses are required, capturing the interaction between these variables and their impact “all things being equal”. Thus, in order to understand more in depth what drove participation, a multinomial logistic regression was conducted. The same predictors as those presented in the cross tabulations were used, in three steps: In the first step, the sociocultural variables, including religious practice, were included; then party proximity was added, and last, attitudes towards immigrants and Islam were entered (Table 1).

Table 1

Explaining the attitudes towards the Charlie demonstrations.

Model 1
Model 2
Model 3
participated vs. refused wish vs. refused participated vs. refused wish vs. refused participated vs. refused wish vs. refused

Practicing Catholics .66 .48* .75 .49t .68 .47*
Occasionally practicing Catholics 1.38 1.21 1.53 1.27 1.47 1.24
Other religion .90 .62 1.00 .69 .89 .65
Muslims .75 .54 .54 .46t .47 .41*
No religion 1.40t .92 1.25 .93 1.09 .88
Non-practicing Catholics (ref)
Birth cohorts
Born before 1940 .55 1.26 .63 1.31 .62 1.34
1956–66 1.49 1.33 1.67t 1.44 1.60 1.38
1967–76 1.49 1.61t 1.83* 1.87* 1.74t 1.81**
Born in 1977 or after 1.47 1.76* 1.78* 2.08** 1.76t 2.01***
1940–55 (ref)
Level of education
Secondary education 1.97* 1.19 2.09* 1.22 2.06* 1.21
Baccalaureate 3.22** 1.36 3.28** 1.35 2.93** 1.29
University degree 5.11*** 1.32 4.17*** 1.14 3.61*** 1.07
Primary education (ref)
Shopkeepers or craftsmen .41 .81 .54 .82 .47 .78
Intermediary professions .88 1.29 .73 1.17 .72 1.16
Employees 1.16 1.57t 1.05 1.49t 1.03 1.47
Blue collars 1.05 1.18 .91 1.13 .86 1.11
No profession .96 1.16 .89 1.13 .82 1.12
Professionals (ref)
Place of living
Suburbs of big cities .55** 1.11 .53** 1.13 .55** 1.12
Medium-size or small cities .57* 1.11 .65 1.25 .68 1.24
Big cities (ref)
Men 1.48* 1.15 1.49* 1.21 1.51* 1.22
Women (ref)
Partisan Proximity
Radical left 4.88*** 1.64 3.53** 1.39
Left (Socialist party) 3.90*** 1.98* 3.15** 1.78*
Centre 1.83 1.19 1.58 1.10
Radical right (FN) .76 .72 .83 .78
No party proximity .92 .60** .76 .58*
Right (UMP) (ref)
Tolerance toward immigrant 1.52** 1.25t
Tolerance toward Muslims .87 .91
Constant .25*** .48*** .16*** .48* .23*** .44t
N 1013 1013 1013
R2 6 percent 9 percent 10 percent

NB: The model is a multinomial logistic regression. Entries are odds ratios: an odds ratio of 1 means that between the tested group and the reference group the chances are even to give a specific answer, an odds ratio greater than 1 that the tested group has more chance to give a specific answer than the reference group and an odds ratio less than 1 that the tested group has less chance to give a specific answer than the reference group.

***p < .001, **p < .01, *p < .05 et t p < .10.

The results confirm our starting hypotheses. The participants to the Charlie rallies were very similar to the usual demonstrators on post materialist issues (van Aelst & Walgrave, 2001; Fillieule & Tartakowsky, 2012). The four variables with the strongest impact on taking part in the demonstrations were, as expected, party proximity, education, birth cohort and positive attitudes towards migrants. Participation, effective or wished, increases among left wing, educated, young and tolerant respondents. In the complete model (model 3) including all the variables, the chances (measured by the odds ratios) to have declared taking part in the demonstrations are 3.6 times higher among respondents with an university degree than among those who stopped after primary education, and 3.5 higher also among the radical left respondents than among those close to the right. As for not taking part, but wishing one had, the birth cohort is a particularly strong predictor. The chances to have given such an answer are twice as high among the younger generations, born after 1977, than among the older ones born between 1940 and 1955. This result do not support Todd’s assumption (2015) that the participants to the Charlie rallies were ageing baby-boomers, eager to preserve their way of life.

Two other variables have a statistically significant but smaller impact on participation: gender and place of residence. Contrary to what was suggested in Figure 1, based on simple cross tabulations, men are more likely than women (odds of 1.5) to have marched for Charlie on January 11. This gap is consistent with previous findings on demonstrations (Mayer, 2010: 223). After controlling for age, education, religion, political leaning, women are usually less inclined to take part in demonstrations, which has previously been interpreted by women negative attitude toward the crowd, the risks of violence and physical confrontation associated with it, in line with gender socialization. Recent comparative research shows a greater reluctance of women for collective and possibly confrontational forms of political participation in general (rallies, meetings), and conversely a stronger inclination to private oriented forms such as boycotts, boycotts or petitions (Coffé & Bolzendahl, 2010). As for residence, as expected, living in a big city boosts the probabilities to have taken part in these rallies, while on the contrary living in suburbs, mostly neighborhoods with higher level of unemployment and poverty, and bad connections with the city center, divides them by two (odds of .55).

Finally, once controlled by all the other individual characteristics, the influence of religion is both less important and more complex than expected. Being a Catholic, whatever the level of practice has no statistically significant effect on the actual participation in the rallies, in all three models (Table 1). It only influenced the wish to have participated, a more tenuous opinion. Practicing Catholics have half less chances than non-practicing ones to say they did not go but wished they had (odds close to .50 in the three models). Islam had more clear-cut effects. Indeed, once controlled for their socio-economic status (lower than average) and their political leaning (more on the left than the rest of the sample), Muslims did not seem to feel quite at ease about the Charlie demonstrations. Compared to non-practicing Catholics, they have half less chances to have taken part in the rallies or wished they had (model 3).


On the whole, our results contradict Todd’s assumptions (Todd, 2015). The “Charlie” in the streets were overrepresented among young, urban, educated, leftwing, and tolerant citizens, i.e. they had the usual profile of demonstrators mobilized on post-materialist issues. They had no hostility to Muslims or minorities, on the contrary: The most inclined to join the rallies were those with the lowest scores on our scales of islamophobia and xenophobia. As for religion, once controlled by all the other individual characteristics (age, gender, residence, party proximity, attitudes), being a Catholic, whatever the level of practice, showed no statistically significant effect on the actual participation in the rallies.

Answers to opinion polls depend on the time of the survey, the mode of data collection, the sampling methods, the questions asked, the way they are framed, etc. Our data has its limitations. For instance, it does not include indicators of emotions, that played a central part in the reactions to the Charlie attacks (for a detailed analysis of the contrasted effects of fear and anger see Vasilopoulos, Marcus & Foucault, 2015). Yet our findings are corroborated by another survey, self-administered online, in a larger sample limited to French citizens, conducted earlier, closer to the events, and with a different question, asking if the respondent had taken part in one of the “Republican gatherings” of 10–11 January (Rouban, 2015). The profile of the 22 percent of the sample who answered positively is very similar to the one we found with the CNCDH survey, with the same over representation, among the respondents saying they went to the rallies, of upper middle class, educated, left wing respondents, also less concerned by law and order issues and more trusting in “unknown others”, and above all, more tolerant, more open to immigrants, foreigners and minority religions than the non participants.

Last, one should note that this tolerant mood, after the bloody attacks against Charlie Hebdo and the Hyper Cacher, was not only a characteristic of those who took part in the rallies, it was shared by French society at large. One can see it on the Longitudinal Index of Tolerance constructed by Vincent Tiberj (Mayer, Michelat, Tiberj & Vitale, 2015 and 2016). Based on 69 series of questions asked in the CNCDH annual Barometer on racism (the survey used for this study) since 1990, the index, which varies from 0 (if every respondent gave the prejudiced answer to all the questions) to 100 (if every respondent gave the tolerant answer to every question), summarizes the ups and downs of tolerance toward minorities and migrants (for a contextualisation see Tiberj, 2008; also Cohu, Maisonneuve, & Testé, 2016). After the Charlie Hebdo and Hyper Cacher gunning, far from declining, the index gained 2 points (between the November 2014 and a special post Charlie March 2015 survey), and after the terrorist attacks of the Bataclan and the Stade de France, on 13 November 2015, 5 points (January 2016 survey). In other words, terrorism does not necessarily breed intolerance. The part played by the government and the political class is essential. All depends on how political leaders, media, institutions, “frame” and interpret the event (Chong & Druckman, 2007). In January and in November 2015 as well, it is the appeal to the Republic and the defense of the freedom of expression that were put forward, not the rejection of migrants, nor more specifically, of Muslims (Mayer, Michelat, Tiberj & Vitale, 2016). This state of grace may not last, but the present trends definitely go against Todd’s predictions.