Integration into Europe: Identifying a Muslim Effect1

Claire Adida – Stanford University David Laitin – Stanford University

Marie-Anne Valfort – Paris I, Sorbonne Part I : Motivation and Broad Results A. Introduction On November 11, 2009 President Nicolas Sarkozy fulfilled a campaign promise to promote a new collective understanding of “what it means to be French.” The setting of his passionate speech initiating a national dialogue, La Chapelle en Vercors (Drôme), was well-chosen. The Nazis had burned it down in July 1944 as punishment for being a bastion of the French resistance, and subsequently shot sixteen of its young men whom they had taken as hostages. Where better to raise issues of national solidarity and the treatment of minorities? The President then designated Eric Besson, Minister of Immigration, Integration, National Identity and Cooperative Development to initiate a “Grand Debate on National Identity.” Besson had the perfect ambiguous biography to give complexity to what it means to be French. He was born in Morocco, with a Lebanese mother and a French father, arrived in France thirty-four years ago at the age of seventeen, and while now serving in a government of the Right was a former member of the Socialist Party (1993-2007). Besson set off the debate provocatively, insisting on a set of “shared values” of the French, and noting their republicanism, their secularism, and their belief in universal values. His Ministry created a website inviting popular participation in this debate, and it has already induced vibrant and highly critical commentary by journalists, scholars, politicians and citizens.2

While the President set the agenda for defining Frenchness quite broadly, the powerful political context of this debate concerns the ability of Europe to assimilate, integrate or even accommodate the post-World War II migration of vast numbers of Muslims. This challenge resonates sharply in France, the EU country with the largest Muslim population.3

1 . This research was funded by the National Science Foundation, Muslim Integration into EU Societies: Comparative Perspectives”, Grant SES-0819635, David Laitin, PI.

Under President Chirac, French policies to incorporate Muslims into a docile and accommodative pressure group had at certain junctures backfired. Under then Minister of Interior Sarkozy, the Conseil Français du Culte Musulman (CFCM) was created in 2005 as a representative body for

2 . Quotations in the paragraph are from Steven Erlanger “France Debates its Identity, But Some Question Why” The New York Times November 29, 2009. The website: http://www.debatidentitenationale.fr/ 3 . While there are no fully reliable census accounts, some estimates (http://www.islamicpopulation.com/Europe/europe_islam.html) put the percentage of Muslims in France about twice as high as the nearest EU countries, Netherlands and Germany, so it has not only the largest Muslim population but the largest Muslim population per capita.

2

Muslims living in France. Yet it quickly turned into an arena of contentious politics: “Although this council was supposed to provide an alternative to foreign interference in French Islam,” John Bowen (2009, 26) notes, “It in fact has had the opposite effect. The Algerian, Moroccan, and Turkish consulates saw the 2003, 2005, and 2008 council elections as opportunities to ratchet up control over their constituents by promoting slates associated with each of the home countries, and they did indeed mobilize these residents of France to vote for their slate.” Many in France worry about the implications of a non-assimilating Muslim population. The Pew surveys4 show for instance that 76 percent of the non-Muslim respondents in France expressed concern over Muslim extremism in their country.5

The apparent failure to incorporate Muslims in France (and other EU states) compared to earlier waves of immigrants may result from two distinct forces (each one perhaps a consequence of the other) that leads part of the French (and other European) host populations to consider Muslim immigrants as a threat. On one hand, the loyalty of Muslim immigrants to their adopted countries is questioned. Using a UK-based survey, Bisin et al. (2008) show that Muslim immigrants do not “secularize” with the time spent in the UK, or at least they seem to do so at a much slower rate than non-Muslim immigrants.6 Moreover, they find that characteristics that are commonly considered as factors of integration (such as high education, high job qualification, living in neighborhoods with low unemployment rate) reverse rather than accelerate the secularization process among Muslims. Using the German Socioeconomic Panel of 2001, Constant et al. (2006) draw similar conclusions. They show that, while being young when entering Germany has a positive effect on assimilation and integration and a negative effect on separation and marginalization among Christian immigrants, this feature has no impact among Muslim immigrants.7 Caldwell (2009) documents the apparent lower loyalty of Muslims to their adopted countries with the low recruitment rates of Muslims into European national armies. In France, although objective analysts such as the International Crisis Group reported no direct connection between the riots in France’s immigrant neighborhoods in 2005 and Islam,8 it was not lost on the general French population that “most of the rioters were of Muslim origin.”9

4 . The Pew Research Center for the People & the Press is a public opinion research organization that studies attitudes toward politics, the press and public policy issues.

In October 2008, spectators whistled derisively during the playing of La Marseillaise before a football game pitting France against Tunisia at the Stade de France. These hisses were popularly

5 . Pew Global Attitudes Project, http://pewglobal.org/reports/display.php?PageID=831. 6 . Their secularization index relies on three questions from the FNSEM (Fourth National Survey of Ethnic Minorities) that was conducted in 1993/1994: the importance of religion; the attitude towards inter-marriage; and the importance of having members of one’s own ethnic group in their children’s schools. 7. The authors give the following definitions: assimilation is “the complete adjustment to or absorption by the host country with cutting all ties to the home country”; integration is “the adjustment to the host country with simultaneous retention of ties to the home country; separation is the fact of “not adjusting to the host country but withholding strong connections to the home country”; marginalization consists in “having only weak attachments to both the host and the home country”. 8 Xavier Ternisien “La France et son islam, vus d’ailleurs” Le Monde, March 11, 2006. 9 Xavier Ternisien “Les ‘barbus’ dans le 9-3” Le Monde, November 17, 2006.

3

interpreted as an insult to the French nation delivered by French Muslims10

and President Sarkozy called for an investigation.

On the other hand, the autochthonous national communities of Europe tend to display hostile feelings against Muslim immigrants. Europe – with states defined by their historic nationalities, all of them in the Christian tradition – has had an historically conflictual relationship with the Islamic world going back to the fall of Constantinople to the Ottomans and the Reconquest of Spain in the 15th century. Thus there is throughout the continent a myth of a “Christian Europe” that reproduces itself despite the continent’s virtually complete secularization in the past century. The myth manifests itself clearly in the application of Turkey into the EU (as opposed to Bulgaria), where suspicions run high.11 It also manifests itself through political parties (for example the Front National in France, which placed second in the presidential elections of 2002), which have mobilized opinion to fight against what could become a Muslim demographic predominance in parts of Europe. The Dutch Freedom Party, representing a coalition of groups threatened by what they see as Muslim values, is gaining rapidly in the polls. In Switzerland, a popular referendum in December 2009 banned the construction of Minarets, an outcome that received sympathetic support from the French President.12

While France has not yet adopted such restrictive measures, many French cities have used a variety of zoning regulations to restrict Mosque construction, which many autochthonous residents fear would, if built, reflect the public manifestation of a permanent Muslim presence in their country (Kepel 1987, 287-312).

Despite the sound and fury, the question whether there is a special Muslim problem for Europe in general and France in particular remains unclear. Several recent studies reveal trends that run against the idea of a lack of loyalty among Muslim minorities to France and against a clear hostility of French people toward them. For instance, the Pew poll of 2006 found that 91% of French Muslims express favorable opinions of Christians, and that they are much more likely to emphasize their national identity over their religious one than are other European Muslims. When asked to choose between religion and nationality as their primary identity, 42% of French Muslims said French first (this was the case of only 7% of British Muslims and 3% of Spanish Muslims). Furthermore, the Pew report claims that “substantial majorities of Muslims living in the European countries surveyed say that [in the two years after bombings in Spain and London, and the Cartoon Crisis in Denmark13

10 . http://news.bbc.co.uk/sport2/hi/football/africa/7671299.stm

] they have not had any personally bad experience

11 . New York Times, September 26, 2007, reports that the EU has officially pictured Europe on its Euro currency that includes (Christian) Belarus, Moldova, and parts of Russia, but not Turkey, which officials admit was stricken from the map. Current French President Nicolas Sarkozy expressly opposes Turkey’s accession into the EU. See Tony Barber. 2009. "Fears grow of Sarkozy initiative to downgrade Turkey's EU bid." The Financial Times (October 15). Available: http://blogs.ft.com/brusselsblog/2009/10/fears-grow-of-sarkozy-initiative-to-downgrade-turkeys-eu-bid/ 12 . http://www.guardian.co.uk/world/2009/dec/08/sarkozy-sympathises-minaret-ban-switzerland 13 . Cartoons in a Danish newspaper that depicted the prophet in an unflattering manner set off a wave of protests throughout the Islamic world as well as crystallized anti-Muslim feelings, to the benefit of a new right party (the Danish People's Party) that evokes anti-Muslim sentiments. See Dan Bilefsky “Cartoon Dispute Prompts Identity Crisis for Liberal Denmark” International Herald Tribune (February 12, 2006).

4

attributable to their race, ethnicity or religion.”14 Following the popular referendum on the construction of Minarets in Switzerland, a CSA survey ordered by the newspaper Le Parisien has shown that a majority of French people (54%) think there is nothing incompatible with Islam being practiced in the French public sphere.15

Although this rate is lower than the one received by Catholicism (82%) and Judaism (72%), the newspaper interprets this result as a signal of a growing acceptance of Islam among French people.

This paper, focusing on the comparative success of Muslim immigrants and their descendants in the French labor market, relies on newly developed social science techniques to adjudicate one element of the “Muslim integration” issue in the current French debate on national identity. We ask: do French Muslims suffer in economic integration in France in ways that would not be so if everything about them were the same, but they weren’t Muslim? To answer this question, in part I, our paper demonstrates that there is an identifiable barrier to Muslim integration. More specifically, it first shows (in Section IB) that previous experimental and survey research on employment discrimination in France conflates the potential effects of religion and country of origin, making it impossible to identify whether French employers discriminate against Muslims qua Muslims. Part I then (in Section IC) introduces a strategy of matching, replicating previous employment discrimination experiments, but in a way that enables us to isolate a possible Muslim effect. The experimental data are unambiguous and show that for at least one job sector, there is significant religious discrimination against Muslims. The next section (Section ID) relies on original survey data that links job discrimination and the economic failure of Muslims compared to matched Christians in France. In Part II of the paper we rely mostly on data from a field experiment conducted in Paris in March 2009 in which we examine a set of mechanisms that can explain our results. More specifically, we find strong evidence of beliefs by French people about Muslims which feed discrimination. We also find evidence that both the rooted French population and more recent Muslim immigrants negatively sanction their own groups for cooperating with the other, and thereby prevent productive cooperation with the other. Furthermore, both these groups, our evidence shows, are less generous to the other the larger the number of members of the other group surrounding them. Finally, we find levels of altruism among our Muslim subjects to be lower than comparable non-Muslim immigrants, especially in their low levels of generosity and expectations toward Muslim women. All of these factors contribute, we infer, to the comparative failure of Muslims to successfully integrate into the French labor market. In Part III, we discuss the implications of our findings and before concluding, address problems that remain unresolved.

14 . For full results, see Pew Global Attitudes Project, already cited. Work by Lawrence and Vaisse (2006, pp. 43-44, 58-9, 66) in France reports similar results. Muslim immigrants, they find, are not all that different from the historic nationalities of European states. In general, they find, the degree of anti-Islamism in police recorded incidents in France is much lower than anti-Semitic ones with a much larger relative Muslim population in France. Those who are Islamophobic tend also to be anti-Semitic and anti-immigrant in general. They conclude, at least for France, that there seems to be no specific anti-Islamic public feeling. See also Simon Kuper “Immigrant Muslims in Belleville”, Financial Times, October 2 2009, http://www.ft.com/cms/s/2/1f4cf7c4-ad5e-11de-9caf-00144feabdc0,dwp_uuid=a712eb94-dc2b-11da-890d-0000779e2340.html. 15 The question was worded as follows: “Selon vous, la pratique de ces religions est-elle compatible ou pas compatible avec la vie en société? 1. Religion catholique ; 2. Religion juive ; 3. Religion musulmane” (Le Parisien, December 10, 2009).

5

B. Previous Research on Employment Discrimination in France among Workers from Immigrant Families CV experiments (also called “correspondence testing”) consist of sending letters from fictitious applicants in response to job advertisements and analyzing the response rate to these applications. The main advantage of correspondence testing is that the experimenter can identify the impact of a small variation in the CV on the response rate, all else held constant. The main drawback is that it only provides a raw measure of discrimination since it focuses on response rates, rather than on actual job offers. In other words, it may underestimate (or overestimate) discrimination in the cases where personal interviews of individuals who were discriminated against during the callback phase yield a lower (or higher) number of actual job offers than personal interviews of individuals who were advantaged during the callback phase. However, prolonging the experiment up to the stage of personal interviews (“audit testing”) would generate further biases in the measure of discrimination because experimental controls would be severely weakened. Even if both fictitious candidates were played by the same person during the phase of interviews (which could be done for religion and not for race or sex), one can still never be sure that the candidate’s performance in each role would be exactly the same.

A wide range of correspondence testing has been conducted throughout the world, starting with the experiment of Jowell and Prescott-Clarke (1970), which tested whether applicants from Asian backgrounds were discriminated against in the UK labor market. To our knowledge, only two of these studies focus on discrimination against applicants from a Muslim background in the French labor market (Amadieu (2004) and Duget et al. (2008)). Both of these studies consist of comparing the response rate received by a CV with a Moroccan-sounding first and last name with the response rate received by a CV with a French-sounding first and last name, all other characteristics being equal across these CVs. Both studies conclude that applicants from a Moroccan background are strongly discriminated against in the French labor market, compared to those with a typically French name. For instance, Duget et al. (2008) compute that a response rate of 100 for the candidate with a French-sounding name corresponds to a response rate of 35 for the Moroccan candidate (and this difference is statistically significant at the 99% confidence level).

Although these studies reveal discrimination of high magnitude against applicants of

Moroccan background, they do not allow us to isolate the source of this discrimination. Two confounding factors are at stake: do employers discriminate against Maghrebis or against Muslims? The difficulties in identifying a religious effect as demonstrated in the CV experiments performed up till now are not easily resolved, and this is all the more so in surveys. Data on Muslims in France are hard to get. A 1978 law set prohibitions on the collection of data on the racial, religious, or ethnic identity of its citizens, creating challenges for demographic research. For instance, in a leading sociological study of the economic success of different immigrant groups, researchers could not distinguish the children of Algerian migrants into France from the children of the pieds noirs, those of European ancestry who left after Algerian independence of

6

1962 (Meurs et al., 2006, pp. 675-76). Although the law was partially relaxed in 2007, this type of data collection has remained stringently limited.16

Equally important for the problem of statistical analysis, those available mass surveys exempt from state oversight rarely (with Pew the partial exception) include enough Muslims to allow for good data analysis. For example, the World Values Survey included in its first three waves in France only 0.4 per cent Muslims, quite unrepresentative of a group that makes up an estimated 6.3 per cent of the resident population.17

But the real killer for identifying a Muslim effect is that most Muslim immigrants to each of the major European states come from a single country or world region. In Germany, nearly all Muslims are from Anatolia; in the UK, the same is the case with South Asians; and in France, nearly all Muslims are from the Maghreb. In each of these cases, it is nearly impossible to determine conclusively whether any additional problems these immigrant populations and their descendents have had in economic mobility beyond the problems faced by non-Muslim immigrant families, are due to the fact that they are Muslims. For example, nearly all immigrants to France from Portugal are Catholic and from Algeria Muslim: once they control for homeland, statistical models cannot distinguish a religion from a country effect on outcomes. Or, to look at the inferential problem another way, any special problems in economic advance faced by the children of North African immigrants to France may be due to their Muslim religion, the fact that 16 . Under Article 8 of the French Data Protection & Liberties Act (Loi informatique et libertés) of 1978, the Commission nationale de l’informatique et des libertés (CNIL) was created as an independent administrative state authority to ensure that outside of the national statistics agency (INSEE), “personal data revealing directly or indirectly the racial or ethnic origin of individuals” could not be processed [by government personnel, or on government contract] without consent. For the law itself, see http://www.cnil.fr/fileadmin/documents/approfondir/textes/CNIL-78-17_definitive-annotee.pdf. For more information on data constraints in France, see the CNIL web page http://www.cnil.fr/english/news-and-events/measuring-diversity/?tx_indexedsearch[ext]=1&tx_indexedsearch[sword]=Constituional+Council&x=24&y=10. Eight exemptions from that prohibition are listed in article 8 of the law, enabling the PRI project, to be discussed shortly, to proceed. For the 2007 CNIL recommendations for relaxation of the 1978 law, see Commission nationale de l’informatique et des libertés (CNIL) (2007). Mesure de la diversité et protection des données personnelles, rapport, 15 mai : http://www.cnil.fr/fileadmin/documents/approfondir/dossier/diversite/RapportdiversiteVD.pdf]. In 2009, Yazid Sabeg, the new Commissaire à la diversité et à l’égalité des chances, with the full support of the president, has taken a new tack on the collection of data on the ethnic and religious self-designation of French citizens, which may prove to be a major breakthrough in future sociological analysis of the population. See Le Monde 18 avril 2009, “Horizons/Débats”, pp. 18-19, “Statistiques ethniques: pour ou contre?” In these debates, Eric Fassin, a sociologist at the École Normale Supérieure explains that the real question is to know what use will be made of these measures. Stéphane Jugnot, a statistician and economist, offers a polemic against Patrick Simon, the leading advocate for the collection of state data on ethnicity, working in the INED [Institut national d’études démographiques], calling his proposals hypocritical, in that the categories of diversity will be racial and will lead to pressures for affirmative action that will reify racial consciousness in the population. Yazid Sabeg has promoted self-attribution. Jugnot claims that this is also hypocritical “because self-attribution only exists if the question of the origins of identity is an open-ended question.” But then, Jugnot predicts, the answers will be so diverse (someone could answer that he is “somewhat Egyptian, somewhat Arab, somewhat Copt, somewhat French”) that researchers, to make any sense of the data, will have to impose a categorization that will reify race and religion in defiance of republican ideals. 17 . The World Values Survey is a global network of social scientists who have surveyed the basic values and beliefs of the publics of more than 80 societies, on all six inhabited continents. Their most recent wave includes 9.4% Muslims in France, but with a sample size of 500, it remains difficult to isolate a Muslim effect with standard controls. See their website at: http://www.worldvaluessurvey.org/.

7

they were a colonized people where in Algeria an insurgency fought against French imperialism, or some aspects of North African life distinct from Islam (the language, the educational systems and the history of authoritarian political rule in the country of their ancestors).

To illustrate this inferential problem, we draw from a large-n survey of 12,010 randomly selected households with an immigrant then (2002-2003) living in France, one of the few large-n surveys in France where a self-reported question on religion was posed.18

The survey also contains key information on the age, sex, education, country of origin and income of immigrant respondents. These data were collected to study the retirement decisions of immigrants, but they allow us to illustrate the problem of trying to infer the effects of religion on income when there is a high correlation between homeland and religion.

Consider Table I-1 that analyzes these data, where Models (1) and (2) illustrate the problem. Model (1) estimates the respondent’s income as a function of his or her religion, controlling for sex, age, education and length of stay in France. With dummies for all religions (and Christianity the omitted category, and thereby serving as the point of comparison), we find that Muslim immigrants are significantly poorer than Christian immigrants. In Model (2), we add home-country fixed effects, a statistical technique that allows us to control for the immigrant’s homeland. The Muslim effect previously identified in Model (1) loses statistical significance. Once we control for home country, we are unable to identify a Muslim effect. If the goal is to identify an independent Muslim effect on an immigrant’s economic integration, this study (as with the previously reviewed CV studies) yields inconclusive results. A different strategy is needed to address our questions.

In response to the confounding effects of religion and homeland in previous studies, we

rely on a social science technique that will allow us to identify whether there exists a Muslim-based discrimination in the French labor market, holding homeland-based discrimination constant. C. Our Matching Strategy Our approach – called “matching” in statistical analysis -- requires the choice of immigrant groups that are divided religiously, with one portion of them being Muslim and another portion, quite similar to the first culturally, economically, and in education, Christian. Comparing the Muslims and Christians in each group allows one to measure the “Muslim effect” without other confounding factors, such as country of origin. 18 . Wolff et al (2007) The PRI [Passage à la Retraite des Immigrés] project was carried out under the direction of Claudine Attias-Donfut, in collaboration with Rémi Gallou and Alain Rozenkier, with funding from the Agence nationale pour la cohésion sociale et l’égalité des chances (ACSE), the AGIRC-ARRCO, the MSA and the Caisse des Mines. This PRI project, completed in 2003 by the Caisse Nationale d’Assurance Vieillesse and the Institut National de la Statistique et des Etudes Economiques (Insee), examines the factors and mechanisms that characterize immigrants’ transition into retirement. Respondents were randomly selected from the Insee Census of 1999, from the baseline population of households with at least one immigrant member between the ages of 45 and 70 at the time of the survey administration. The resulting sample comprises 6,211 respondents, 46.4% women, of mean 55.8 years and median 55 years of age. It is representative of the immigrant population residing in metropolitan France in 2003, that is to say, of all foreign-born immigrants in the selected age range. While the data remain private, the authors kindly provided the data permitting our analysis in Table I-1.

8

Proper matching requires that the two religious subsets of religiously-divided immigrant groups arrive at the host country with relatively equal resources, or else the comparison would be biased. Suppose the target population were Lebanese in Western Europe or North America, and the comparison were between Maronite Christians and Sunni Muslims. Because the Maronites start off earlier and with rich international networks of banking families, a finding that Maronites achieve higher rates of economic and social success would tell us little about comparative economic barriers in the West due to religion. Careful examination of selected small immigrant groups in France, however, invites opportunities to get a reasonable approximation to an unbiased comparison. In our case, we identified an estimated 10,000 immigrants in France with family backgrounds as Joolas and Serers, two distinct ethno-linguistic communities from Senegal (hereafter, the Serer and Joola Muslims from Senegal will be called SM’s; the Serer and Joola Christians from Senegal SX’s).19 These two groups, unlike all other communities in Senegal, have a sufficiently large Christian population to allow for intra-group comparisons.20 Moreover, contrary to what we observe for Lebanese Maronite Christians and Sunni Muslims, SX’s did not benefit from the earlier settlement of a Senegalese Christian diaspora in France. To confirm this, we rely on a survey administered by David Laitin with the French firm CSA, in which 511 Serers and Joolas, both Christians and Muslims were polled on a variety of issues concerning their integration in France. Data from that survey indicate that the time elapsed since the settlement of the first migrant is 39 years for Senegalese Christian families and 39.3 years for Senegalese Muslim families. This difference is not statistically significant.21

From this choice of Senegalese Muslims vs. Christians, David Laitin in collaboration with the firm ISM-CORUM was able to replicate past CV experiments in a way that allows us to identify whether there is a religious element in job discrimination.22

19 . We include Manjaks, a closely related linguistic group, with the Joolas. See Summer Institute of Languages Ethnologue, http://www.ethnologue.com/show_country.asp?name=SN.

Keeping with the matching strategy outlined above, in order to separate out region of origin and religion, our experimental design demanded three comparable CVs – two of them from women with an obvious Senegalese surname (Diouf), but one of whom had a well-known Muslim first name (Khadija) while the

20 . From the 2002 Senegalese census, 25% of the Joolas and 11% of the Serers are Christian, while only 5% of the Senegalese population as a whole is Christian. For the latter figure, see http://www.adherents.com/adhloc/Wh_295.html#677. 21 . This survey was conducted in 2009 under contract by CSA France, in a project in which David Laitin, Yann Algan, and Vincent Tiberj are principal investigators. There were 511 respondents from Serer or Joola backgrounds (with 509 giving a clear indication whether their household is Muslim or Christian). The principal sample (n=332) was through lists of mobile phone numbers in France associated with known Serer or Joola surnames; a secondary sample were through face-to-face interviews (n=179) in three regions (Iles de France, Provence/Alpes/Côte d’Azur, and Rhone Alpes) regions, relying on chain referrals. The survey had 29% Christian respondents and 71% Muslim respondents. Descriptive statistics of the survey are in Table A-1 in the Appendix. Hereafter this survey will be referred to as the Laitin/CSA survey. 22 . ISM-CORUM had already administered correspondence tests. See, for instance, “Résultats du Testing Sollicité par le Groupe Casino : Un diagnostic partagé sur les discriminations liées à « l'origine, » Fabrice Foroni avec les contributions de : Eric Cédiey (accompagnement scientifique), Amandine LUC (analyse statistique), Sandrine Argant, Raphaël Bosch et Florence Marfaing (réalisation des tests) Juin 2008 (ISM-CORUM).

9

other had a well-known Catholic first name (Marie); the third CV was from a woman with a typical French name, with minimal religious connotation (Aurélie Ménard). The basic qualifications and background of the three applicants, all French citizens with two years of post-secondary education, unmarried and with three years experience on the job market to reflect experience in the advertised job sector, were the same.23 But here we introduced two additional reinforcing signals of religious identification. One of Khadija’s past positions was with « Secours Islamique »; one of Marie’s was with « Secours Catholique »; while Aurélie worked solely in secular firms. Also, Khadija did voluntary work for the Scouts Musulmans de France, while Marie did the same for the comparable Catholic organization, Scouts et Guides de France.24

Responding to posts on the ANPE (the French national employment agency) website or ads in newspapers, ISM-CORUM administrators sent CVs to all jobs listed for assistante de direction (executive secretary), aide-comptable (junior accountant) or related jobs. We chose these two general sectors because the jobs entailed some exposure to each firm’s clients and customers, jobs in which the human resources division might be especially concerned about how the public might react to the ethnicity or religion of its employees. Half of the firms that posted job positions received applications from Khadija and Aurélie; the remaining half received applications from Marie and Aurélie. For each job that received applications from Marie/Aurélie, the ISM-CORUM team administering the tests sought the closest job offering to send applications for Khadija/Aurélie. In the Appendix, we provide examples of the three CVs. Using what is called a « difference-in-difference » design, our analysis compared how well Khadija’s and Marie’s applications were received, based on how each compared to Aurélie’s in engendering an interview from the employers who advertised for these jobs.25

The results are clear. The two sets of employers were consistently favorable to Aurélie, but strikingly more so in her match-up against Khadidja. Our test candidates, Marie and Khadidja, fared significantly differently. Marie received call-backs at a rate of 21%; compared to 8% for Khadidja, both in competition with Aurélie, who did equally well in competition with both of the Diouf applications. This is more than two and a half times better. The difference in difference test is significant at the 99.99% confidence level.

In sum, our experiment allows us to see, controlling for ethnic origins, whether there is religious discrimination (aimed at Muslims) in the French labor market. See Table I-2a for a summary of the results.

23 . The first 214 job responses had no pictures of the applicants. The next 50 had the same picture for the two Dioufs, as a signal that they were not North Africans. There were no religious symbols in the photos. The results were not statistically different and in this paper both sets are combined. 24 . The first two organizations are both independent NGOs in France with extensive international networks dedicated to the eradication of poverty. Neither has an obvious religious agenda. See their websites: http://www.secours-catholique.org/ and http://www.secours-islamique.org. As for the volunteer work, the two organizations are listed as religiously affiliated branches of an international scouting federation. See http://www.sgdf.fr/-Connaitre-le-mouvement-. 25 . These tests were performed from March through September 2009, during a period of grave economic crisis. Few jobs were advertised, and the overall rate of positive response for a first call-back was only 21 percent.

10

We then analyzed the likelihood that our control (Aurélie) and our test (Marie/Khadija) candidates received a call back based on whether the employer received the Aurélie/Marie pair or the Aurélie/Khadija pair and controlling for additional factors such as the type of position, the type of employment contract, the size of the recruiting firm and whether the CV included a candidate photograph. We specified a logit regression (call back or no) with robust standard errors (See Table I-2b). We find that when our test candidate is Khadija (vs. Marie), the likelihood that Aurélie gets a call back increases and the likelihood that the test candidate gets a call back decreases. This result is significant at the 95% confidence level.

Controlling for ethnicity, this experiment provides a clear indication that in at least one sector of the French labor market, there is significant religious discrimination.

D. Consequences of Religious Discrimination This section suggests a link between job discrimination and the economic returns for Muslims in France. We rely on the Laitin/CSA survey to compare income levels of Senegalese Christians (SX’s) and Senegalese Muslims (SM’s) living in France in 2009. We previously established that these two groups immigrated into France in a single wave during the 1970s. To ensure a fair comparison, we must also ask whether the first immigrants of respondent families to France started out on equal footing. To address this concern, we relied on evidence from the most recent Senegalese census. The only critical difference (that might have proven to be an advantage for SX’s) upon the arrival to France relates to education: SX’s were slightly more educated than SM’s: while the probability of having a secondary or a post-secondary education is 36% among SX’s, it is 27% among SM’s (this difference is significant at the 95% level).26 These results are consistent with ethnographic accounts of Senegalese Christians’ access to better quality education through their religious network (i.e., Catholic schools).27

A proper matching strategy between SX’s and SM’s thus requires that we control for the first migrant’s level of education.

We can now ask, controlling for the education level of the first migrant, whether SX’s have been more successful in breaking through the economic glass ceiling in France than have SM’s, and if so, why? With this procedure, we are confident that any differences found between the two groups are the result of some aspect of their religious upbringing or practice, or the way they are perceived by French people, since geographic origin does not vary and initial human capital is controlled for.

26 . Data from the 2002 Senegalese census, which we will report on in a future paper, reveal precisely the same degree of educational difference between the subset of Muslim and Christian Joolas and Serers who have a relative living in Europe. 57% of Muslim respondents who had a relative in Europe had only a primary or middle school education, while 42% had higher levels of education. For Christians, the figures are 53% primary or middle, and 47% more advanced. These differences are significant at the 99% level, but substantively are not radically different. These data add confidence that the Laitin/CSA survey was representative of the two ethno-linguistic migrant groups in France. Thanks to Chris Beauchemin for giving us access to the census data, to Susan Holmes for technical assistance, and to Jessica Gottlieb for research assistance. 27 . We owe our interpretations of Senegalese culture and society to our field collaborator, Etienne Smith, who has conducted extensive field research in Senegal.

11

One might ask if choosing Black Africans as the experimental targets could capture any religious effect in France. Indeed, there is a myth in France, held particularly strongly by Arabs that African Muslims are not really Muslims, since they don’t speak Arabic and mix freely with non-Muslim Africans (Diop 1988). To the extent that its acceptance is widespread, however, this myth could weaken any Muslim effect, but the bias would be against finding such an effect. Ours is thus a more demanding test for identifying a Muslim effect.

We estimate the determinants of immigrant income in France today via an ordered probit

regression with robust standard errors, with the current monthly household income as the dependent variable. The explanatory variables are the religious tradition of the household, the head of household’s gender, and ethnicity and the education level and year of immigration of the head of household’s ancestor who was the first to come to France. The results are reported in Table I-3. (Recall there is no need to include country fixed effects, as all respondents are from Senegal. This is the key to our matching strategy). We find that households with a Christian religious tradition are significantly richer than households with a Muslim religious tradition (significant at the 99% level). More precisely, the probability of having a household income greater than the median in our regression sample (N=312) decreases by more than 25% when the household is Muslim: this probability is equal to 52.0% when the household is Christian and to 37.9% when the household is Muslim. In sum, even controlling for the education level of the first migrant to France, there is a significant negative Muslim effect on present day household income. We can therefore infer that the job discrimination revealed in Section IC has broad implications for differences in today’s household income for Muslims. Part II: Search for Channels through Field Experiments

Part I demonstrated that controlling for country of origin, ethnicity and initial human capital, the descendants of Muslim immigrants into France have faced higher barriers to economic mobility. But the precise mechanisms driving these results are ambiguous and subject to diverse interpretations. To help resolve these interpretations, we organized a set of field experiments that we now describe. In the following section, we outline our principal findings.

A. Procedures and Protocols

We supervised eight sessions of games held in a rented private language school in the

diverse setting of Paris’ 19th district. Each session comprised a minimum of ten players, and a theoretical maximum of 15, though in practice the largest session had 14 players. Three of the sessions had all women players; three had all men; and two were mixed gender. In these sessions, we conducted a set of experiments in the game theory tradition (Camerer 2003). We elaborate on key aspects of our protocols below. The setting. Our identification strategy required us to embed SM’s and SX’s (whom we refer to as our target population) in a context that would seem natural to them, devoid of any signal that we were seeking to isolate the effect of religion on behavior. We chose then to conduct the experiments in a private language school in the heart of an ethnically diverse district of Paris, the 19th. In the 19th district (compared to the figure for all Paris), the average size of a household is 2.15 (1.87); the percentage of adults who are manual workers is 20.9 (14.5), the percentage

12

living in social housing is 40.8 (19.7), and the percentage born in France is 63.5 (82.4).28 In this significantly immigrant district, for subjects to see a few Africans planted into their game sessions was hardly eyebrow raising. Indeed, our solution to the hiding of our identification strategy worked. In exit surveys, not a single subject speculated that religion had anything to do with the purposes of the games,29

and only one of the target players out of a total 29 verbally wondered if there was something odd about having other players in the room who were from the same language group as they were in Senegal. This player never speculated about religion.

Random selection procedures. For the non-targeted players, we used a stratified (by population density) but not fully random recruitment procedure centered on the twenty-one metro stations in the district. In a fully random protocol, we assigned a weight to each metro station based on the density of the area in which it is located, with the higher density stations getting more cards in our random draw. Each recruitment team would draw a metro station for each recruitment day, and then a number from 1 to 10 to determine which passer-by to invite as game recruit. But because we wanted to have a large number of interactions between our targeted sample and the rooted French population, we deviated from this protocol to assure ourselves a sufficient number of players who would locally be considered authentically French – i.e. with four grandparents born inside metropolitan France. In politically incorrect terms, these French are called “français de souche”, and we shall refer to them subsequently as FdS. When potential subjects who looked as if they were FdS walked by, recruiters were instructed to ignore the sequence of selection, and to ask them if they were interested in participating in our experiment.

Potential subjects who were willing to hear our appeal were told that they could win up to 148 Euros for about two and a half hours of game participation, games which were designed to learn how people from Paris and its surroundings made decisions about money. Turn-downs were about 30 percent, introducing some biases that have no easy interpretation.30

Those who accepted needed to register themselves during open enrollment periods at the language school. At registration, they filled in a personal questionnaire that collected socio-demographic information, which we used for purposes of staffing the game sessions. They also played a low-stakes altruism game advertised to them as practice for the kinds of games they would play a couple of weeks later during the actual sessions. Just for coming to registration, players were given 15 Euros (approximately $22 at that time) plus transport money. We enrolled sixty-three non-targeted players; 73 percent were born in France (this is higher than the average for the district, but lower than for Paris generally).

The targets. The protocols called for three target players (two from one religion and one from the other) for each session. Of our 29 targets, 18 self-identified as Muslim, 10 as Christian, and one 28 . Data supplied by the mayor’s office at the 19th district. On foreign born in Paris, see http://www.migrationinformation.org/dataHub/GCMM/Parisdatasheet.pdf. A good picture of the diversity in the 19th district is offered in the French film “Entre les murs” (“The Class” in its English-language version). 29 . In the exit questionnaire, we asked : “Selon vous, quel était le but de cette étude?” [What was in your opinion the goal of this study?] 30 . Orthodox Jews turned us down, not only on Saturday recruitment days when they were prohibited from writing, but other days as well. We did not successfully recruit any Asian players, even though by observation (but not in the town data) there were many Asians who were present in the neighborhood.

13

with neither world religion. We relied upon three separate networks to recruit these players. Two of the networks came from the ethnographers who were conducting family histories for our wider research project. Our ethnographers were asked to recruit subjects by merely telling them they had heard about these experiments with a chance to earn a lot of money. No mention was to be made about Senegalese specificity or religion having anything to do with the games. The third network came from a Senegalese night watchman (not of the target populations) who worked at a student dorm. He was given a quota for the targets and paid for each recruit who showed up for inscription and participated in the games. As feasible as possible, each session combined one target from each network, to avoid pairing up players who knew each other. When not feasible, we relied on a local informant who advised us on which of the Senegalese were most likely not to know each other. We asked in the exit survey if any of the players knew another player at their session, and our analysis accounts for dyads in which players reported knowing the other beforehand. In Appendix Table A-2, we perform a balance test that assures us that the two sub-samples, SM’s and SX’s, are similar on basic socio-demographic variables, differing significantly only in their religiosity and their stated level of trust. We therefore control for these two variables in our analysis (this will be done in the next iteration of this paper). At the Game Site: Based on information learned at registration, subjects were assigned to a session that would last about two and a half hours. When they arrived, they were assigned a code number that allowed us to organize our SX, SM, and FdS players in a way that would maximize their interactions. They were also asked to write their first names on a label and to paste that label on their chests. The only information players had about each other was their looks, their manners, their dress and their first names. After check-in was completed, they were brought to an open room with the other players, given magazines to read, and monitored in a way that discouraged any conversation or interaction. Throughout the session, and out of sight from the actual players, monitors kept a full account of all answers and earnings for each player. At the end of the session, as players answered an exit survey, the winnings for each player were placed in sealed envelopes for them to take home. None asked for, nor were any of them told, how much they won from each of the particular games in the protocol, meaning that players never found out how they performed in each game. The Protocols: Organized by sequence, Table II-1 outlines key information for the seven games (or elements of games) in our protocol. This Table should be used as a guide as we take readers seamlessly through the variety of experimental results. B. Hypotheses and Experimental Results We hypothesize three channels through which the Muslim disadvantage might flow: tastes, beliefs and technologies.31

Below are the hypotheses to be explored on each of these channels.

31 . Previous experimental work has identified three principal channels in which ethnic diversity could promote market inefficiencies: preferences for discrimination, technologies of communication, and social sanctioning raising the costs of in-group defection (Habyarimana et al. 2009). With a different dependent variable from theirs (ours being relative economic success; theirs being collective community action), our theoretical conjectures called for a different categorization. Our categorization served to help organize the patterns we observed ex post; we did not theorize this set of channels ex ante.

14

Tastes: H1 – The FdS have anti-Muslim tastes that lead them to pay a cost for their discriminatory behavior H2 – The SM’s do not have a taste to invest in cooperative relations with any societal groups H3 – The SM’s have a taste for discrimination against their own women H4 – The FdS feel more uncomfortable the more SM’s are around them, which is not the case for SX’s H5 – The SM’s feel more uncomfortable the more FdS are around them, which is not the case for SX’s Beliefs: H6 – The FdS believe either that they can make better judgments about Christians than about Muslims or that they have prejudicial beliefs about Muslims in general. In either case, they discriminate. (This is not taste-based, but rather statistical discrimination). H7 – The FdS believe that other FdS have a distaste for Muslims and will sanction other FdS if they cooperate with Muslims, and therefore are less cooperative toward Muslims when there are more FdS around them H8 – The SM’s believe that other SM’s have a distaste for FdS, and will sanction other SM’s if they cooperate with FdS, and therefore are less cooperative toward FdS when there are more SM’s around them. Technologies: H9 – SX’s through shared networks (at Church) with FdS are better able to coordinate with FdS than SM’s, thereby yielding higher levels of cooperation Our strongest findings are: (1) FdS exhibit statistical discrimination against SM’s (Hypothesis H6). Conversely, we have found no convincing evidence that this is complemented with taste-based discrimination (H1);32

(2) SM’s do not have a taste for cooperation with outsiders, especially when stakes are high (Hypothesis H2); (3) SM’s exhibit a taste for discrimination against their own women (H3); (4) There is a strong sanctioning mechanism by FdS’s against FdS’s who try to cooperate with SM’s and a strong sanctioning mechanism against SM’s by SM’s who try to cooperate with FdS’s (Hypotheses H7 and H8) (5) There is mutual discomfort between SM’s and FdS’s, which is not exhibited between SX’s and FdS’s (Hypotheses H4 and H5); (6) There is no evidence for SX’s having a more efficient technology of communication with FdS than the SM’s (Hypothesis H9).

(1) FdS exhibit statistical discrimination against SM’s (Hypothesis H6), but the data do not reveal any pure case of taste-based discrimination (Hypothesis H1): The CV experiment described in Part I is powerful evidence of discrimination by French organizations against Muslim applicants, but doesn’t help discriminate between taste-based and statistical discrimination. Results from the trust game reinforce those we found in the CV experiment, but with the same confound. Consider the receiver’s decision. (Recall that this game is played 32 . We are presently designing an experiment that would give clear evidence on whether there is taste-based discrimination against SM by FdS.

15

simultaneously, meaning that Player 2’s move occurs without any knowledge of Player 1’s decision). Table II-2 shows that FdS, in the position of Player 2, send back significantly less to SM than to SX (with the difference of means test significant at 90%). This result is robust and more significant (at 95%) with an OLS estimation that controls for players’ socio-demographic characteristics and session fixed effects, and which clusters the standard errors at the individual level. But what mechanisms does the receiver’s return capture? The receiver’s decision has an element of taste (as there is no strategic advantage in rendering any return to the sender and any amount sent could be seen as generosity or good will). But it also reveals (second order) beliefs that you ought to return money to players you believe will be trusting of you. The return by sender therefore adds confidence that there is religious-based discrimination. But from both the CV experiment and the trust game, it is difficult to separate out whether the discrimination is a matter of taste, or whether it is based on beliefs by FdS about the probability of successful exchange with SM’s (given either the generalization of popular prejudices against SM’s, or greater uncertainty about the productivity of SM’s relative to SX’s).The trust game, therefore, doesn’t allow us to discriminate between H1 and H6. The dictator game is the cleanest test of taste. Here we find that FdS donate no more money to SX’s than to SM’s. There is therefore no pure evidence of taste-based discrimination that undermines SM economic advance in France. We capture statistical discrimination in game moves that rely predominantly on the beliefs that players hold about their partners. In the voting game, subjects vote for a leader who will then allocate money to them. The actual allocation has a strong altruistic element. But the vote itself is based largely on trust, that is, on the expectation that the leader will respond with a generous allocation to the voter. It therefore captures beliefs rather than tastes. In this voting game, FdS players are on average less likely to vote for a SM leader (0.14) than for a SX leader (0.27). Although this difference of means is not statistically significant (p=0.16), the OLS coefficient on the effect of the candidate’s religious identity on an FdS’s likelihood of voting for him or her, is statistically significant at the 90% confidence level. Here, we find some evidence suggesting that FdS statistically discriminate against SM’s relative to SX’s. FdS statistical discrimination reveals itself more weakly in two other instances. In the trust game, FdS senders on average allocate 2.43 Euros to SM’s and 2.55 Euros to SX’s, showing stronger beliefs in a higher return from SX’s. While not significant (p=0.28), this difference is consistent with our interpretation of the voting results, that FdS trust SX’s more than they trust SM’s.

Finally, when asked whether, in the speed chatting game, they would recommend players they just met to an employer, FdS are less likely to recommend their SM than their SX interlocutors. This difference of means is significant at the 90% confidence level, but the OLS estimation does not yield significant results. [These inconsistent statistical results need to be further analyzed]. Overall, there is consistent (though not always significant) evidence of greater FdS trust toward SX’s than toward SM’s.

The experimental results support a statistical foundation for FdS discrimination, one

based on beliefs rather than tastes.

16

(2) The SM’s fail to establish cooperative arrangements with any groups when stakes are high (Hypothesis H2): Cooperative arrangements, we assume, are reached when players can overcome their rational, egoist tendencies and exhibit altruism toward others. Here we find that SM’s are much less altruistic (keeping what they can for themselves) that SX’s when the stakes are worth keeping.

Table II-3 illustrates differing levels of altruism among FdS’s, SM’s and SX’s across games, ordered from games with low to high stakes. In the practice altruism game where the stakes are very low (3 Euros), we do not find that SM altruism is significantly different from SX altruism. In fact SM’s give more, though not significantly so. In the trust game, where the stakes are bit higher (9 Euros), SM’s return a lower share to all senders than do FdS or SX, although this is only significant relative to what the FdS return. Once we reach the two high-stakes games, however (30 Euros), SM’s are significantly less altruistic to all players relative to both SX’s and FdS’s, who do not differ significantly in their level of altruism.

The change in SM behavior is likely due to rising stakes rather than pre and post speed-chatting interactions. Indeed, the change in SM behavior begins with the trust game, which takes place before the speed-chatting interactions. The more convincing explanation is therefore that generosity goes down for SM’s (but not for SX’s) when the stakes go up.33

Furthermore, we note that the low level of altruism toward the majority group is not counterbalanced by in-group generosity, and this might raise the barriers for group advancement.

We have evidence that this is a matter of taste, rather than a fundamental distrust of societal others. In the Laitin/CSA survey, using the standard World Values Survey question on trust, and controlling for ethnicity and income, we find that SM’s are no less trusting of others than are SX’s. On the other hand, we observe in that survey, again with standard controls, that SM’s send back more money and more often to their families in Senegal than do SX’s; and they prefer to be buried back in Senegal far more than do SX’s. This suggests an orientation to family rather than society, and provides the best explanation for low levels of SM altruism to all groups (including other SM’s) compared to SX’s.

(3)The SM’s exhibit a taste for discrimination against their own women (H3): SM altruism reaches the lowest bounds when it comes to allocations to their own women. To test the channel of gender discrimination, we concentrate on the amount given in the dictator game for two reasons: first, the dictator game is most likely to cleanly capture altruism; second, our number of observations is highest in this game, allowing us to differentiate behavior by sex in both difference of means analysis and regression analysis. Table II-4 indicates that both SM males and SM females are significantly less generous to SM females than they are to SM males. In comparison, SX males and females are no less generous to SX females than they are to SX males.

33 Low levels of SM altruism are not significantly different across groups of players (SM’s display similarly low levels of altruism toward SM’s, SX’s and FdS’s). However, as we shall see below, the one group that does suffer from lower levels of SM altruism are SM females.

17

This finding is consistent with much other research revealing a gender bias among Muslim respondents in the Norris/Inglehart (2003) study of World Values, and our CSA survey as well. As shown on Tables II-5a and b, compared to SX’s, SM respondents have lower educational hopes for their daughters, and are more likely to see the fundamental place for women to be in the household. A potential explanation for the results in Part I of our paper is lower SM altruism toward all groups of players (undermining useful cooperation even among themselves) and greater SM taste for discrimination against their own women (lowering human capital). (4) There is a strong sanctioning mechanism by FdS’s against FdS’s who try to cooperate with SM’s, and a strong sanctioning mechanism by SM’s against SM’s who try to cooperate with FdS’s (Hypotheses H7 and H8): Does group sanctioning inhibit or encourage SM integration? Here, we focus on the channel of in-group policing. Namely, we examine whether the presence of more FdS leads any individual FdS to be less generous toward SM’s than toward SX’s, or if the presence of more SM’s makes any individual SM less willing to be generous toward FdS’s than would be the case for SX’s. Tables II-6a and b, relying on data from the dictator game, where the number of SX’s, SM’s, and FdS’s vary across sessions, present the results of our tests for these mechanisms. We find both in-group sanctioning effects that work against inter-group generosity. The Tables reveal, first, that SM’s give less to FdS guises in the dictator game with more of their own group is in the room. This is not the case for SX’s, who do not significantly change their behavior with more SX’s in the room. Additionally, FdS’s give less to SM’s (but not necessarily to SX’s) with more FdS’s in the room. In other words, SM’s seem to sanction their own when it comes to generosity toward FdS; and FdS seem to sanction their own when it comes to generosity toward SM’s. We do not observe this for FdS-SX interactions. These results suggest a norm among FdS’s to punish other FdS’s who are generous to SM’s, and a norm of SM’s to punish other SM’s who are generous to FdS’s. Both of these norms, we surmise, hinder SM integration into France. We find another piece of evidence supporting H7 is in the results from the strategic dictator game, where we can test what beliefs players hold about FdS generosity. Recall that all players were asked to guess the allocations of three people in the dictator game: one of the FdS players who they were told was randomly chosen from the group; and two French-looking women, one young and the other elderly, whose pictures were broadcast on a large screen. Players in general did not guess there would be any discrimination against the Muslim guises as played by these three FdS models. But, as shown in Table II-7, FdS players did infer that the elderly model would reveal anti-Muslim feelings by her allocations. (Alternatively, FdS subjects guessed that the younger FdS model would give more to the Muslim guises). If we can assume that elderly French women are a good proxy for traditional cultural beliefs, then all FdS, when there is in-group policing by bastions of French traditional culture, will strategically adjust their behavior to coincide with their anti-Muslim cultural norm. (5) There is mutual discomfort between SM’s and FdS’s, which is not exhibited between SX’s and FdS’s (Hypotheses H4 and H5): We call out-group salience the strong feelings of discomfort a player might experience when surrounded by groups of “others”, the more so the larger that

18

group. Because SM’s are a double minority in France (Muslims and Africans) while SX’s are only a minority as Africans, the salience of being an out-group may be stronger for the SM’s than for the SX’s, and stronger for FdS’s interacting with SM’s than for FdS’s interacting with SX’s. Such out-group salience may alter SM and FdS behavior (notably by reducing inter-group cooperation) in such a way as to exacerbate anti-Muslim discrimination. The dictator game provides some evidence in our experiments for this out-group salience effect, as illustrated in Table II-8. In sessions in which there were a greater number of FdS players, SM contributions to FdS guises went down while SX contributions to FdS guises went up. Conversely, the more SX’s were in the room, the more FdS gave to the SX’s, but there was no change in FdS giving to SM guises with more SM’s in the room. These results suggest that SM’s are uncomfortable around too many FdS, while FdS are less uncomfortable around more SX’s than they are among more SM’s. Thus the salience of the SM’s as an out-group is greater than for SX’s, with an impact on behavior, both of SM’s and FdS players. (6) There is no evidence for SX’s having a more efficient technology of communication with FdS than the SM’s (Hypothesis H9 no different from null): Do SM’s suffer from a communication disadvantage? We test this channel through the speed chatting game. In France, there is no linguistic advantage for SX’s over SM’s (both have the same language repertoires: fluency in the mother tongue, variable fluency in Wolof, the language of wider communication in Senegal, and at least adequate conversational French). We are therefore examining a more subtle form of communication advantage: whether a common religion involves “network capital”, the ability to connect with people and infer correct things about them from greater exposure to people like them (people in the other player’s social network). To the extent that SX’s would more likely meet FdS’s in Church (and share at least one social network) than would SM’s meet FdS’s in the Mosque, the SX’s in our sample may enjoy more network capital than do SM’s, facilitating their socio-economic advancement in France. Table II-9 displays results from our analysis of the speed chatting game and quiz. It indicates that there is no significant communication advantage in learned characteristics shared between SX’s and FdS’s. SX’s were no better at learning personal characteristics about their FdS interlocutors than were the SM’s; nor were they better able to make correct guesses than SM’s about their FdS interlocutors. FdS players were not significantly better able to learn the personal characteristics of their SX interlocutors than they were of their SM interlocutors. Finally, although FdS’s made significantly more correct guesses about their SM interlocutors than their SX interlocutors, this is fully explained (as we shall shortly see) by FdS more likely guessing that SX’s were Muslims than guessing that SM’s were Christians. In sum, there is no significant evidence of SX’s ability to exploit network capital for strategic advantage among FdS.34

Part III: Interpretive Issues, Extensions, Summary, and Conclusion A. An Interpretive Issue

34 . In one of our ethnographic interviews (a part of the project not developed in this paper), one SX respondent laughed when asked if she had met many FdS at Church. She responded that there weren’t any FdS who go to Church; she only meets those from the Antilles there!

19

One big question we raise in our interpretation of the experimental results is whether the FdS are conditioning their behavior on our targets’ “religion”, and if not religion, what other signals are they relying upon. We know from the speed chatting game that FdS do not easily recognize SXs as Christians. There were 15 FdS/SX meetings in the speed chatting game; in 5 of those cases (33.5%), the FdS thought that their SX interlocutors were Muslims while in only 3 cases (20%) did they correctly identify them as Christians. (1 was identified as an atheist and in six cases the FdS circled “don’t know” in regard to an SX’s religion). Meanwhile, of the 22 FdS/SM meetings, 11 (or 50%) of the FdS correctly thought their interlocutor was a Muslim. (4 or 18% thought they were Christians and 7, or 32% did not know). In other words, FdS did not correctly identify the target population’s religion 59% of the time; they were incorrect 50% of the time for the SM’s and 80% of the time for the SX’s. Furthermore, from observation, neither SM’s nor SX’s appeared at the game site with religious dress or jewelry. These results suggest that our findings relating to FdS behavior toward SM’s and SX’s are not explained by a correct reading of the target’s religion. Instead, the only clear, differentiating signal FdS’s received were players’ first names, which they wore as a nametag throughout the session. Here, Table III-1 indicates clear differences between SM and SX first names: SX names were more recognizably French, while SM names were more recognizably foreign. Only 4 of 17 SM names were common Muslim names. Therefore, it seems that FdS are conditioning their behavior on the degree of exoticness of the first names of their interlocutors, not on their religion. B. Extensions Although we have uncovered mechanisms that account for the variation in the behavior of SX’s and SM’s in regard to FdS and vice versa, we have not explored whether the sources of these differences are exogenous (having to do with the religion of the targets which in this case would be seen as an exogenous treatment) or endogenous (having to do with the nature of the French context in which SM’s and SX’s interact with FdS). One test for exogeneity is to use data sources such as the World Values Survey and Afrobarometer to see if in a variety of countries (and in particular Senegal) Muslims are more family oriented and therefore less altruistic to others.

A good extension would therefore be to conduct similar game experiments with the Joolas in Senegal. Coming from Casamance, all Joolas would be seen as minorities (compared to the Wolofs) in Dakar. But here, since the majority Wolofs are Muslim, the Joola/Christians would be the double minority (ethnic community and religion) while the Joola/Muslims would be only a single minority (ethnic community). If the Joola/Christians in experiments conducted in Dakar acted like the SM’s in Paris, we would have evidence that minority status drives the results; but if the Joola/Muslims conducted themselves in Dakar as did the SM’s in Paris, we could identify the effect as religious. In this paper, we have made assumptions about the recognizability of names (as possibly French or Muslim or foreign). We are in the course of developing lists of the most common names in France, and those in Algeria (as a proxy for perceived Muslim-ness), to put our assumptions about name recognition to test. A further extension would be to interview random

20

subjects in the district in which the experiments were held in order to get local understandings of how the various SM and SX first names would be interpreted. Another extension would be to rely on the same survey outlined above, showing the pictures of each of the players in our trust games, and asking respondents to rank them as to how good looking each player is. We would then have a proper control (“beauty”) for how much the sender gives in the trust game, as suggested in the work of Wilson and Eckel (2006). A final extension would be to inquire as to why Muslims, in the face of discrimination, do not change their exotic (to the FdS) first names to ones that would be more readable to the FdS. We conjecture this has much to do with SM’s retaining greater connections with the home country, and fearing reprisals there if they return home with Christian-sounding names. This conjecture can be put to test with data on the amount of money respondents send back to their families in Senegal, and their expressed preferences for burial in Senegal. On both of these variables, preliminary tests relying on the Laitin/CSA survey support our conjecture. C. Summary and Conclusions In this paper we have shown that both public debate and previous research on the implications of religious difference for integration into France have yielded ambiguous results. Relying on a new matching strategy, and replicating well-understood CV experiments, we have been able to identify a statistically and substantively strong level of religious discrimination in at least one sector of the French labor market. Relying on that same matching strategy, but now through a large-n survey of the descendents of Senegalese migrants into France, we are able to show one implication of job discrimination, namely that over generations Muslims have done less well economically than have comparable Christians. Part I of our paper established a clear, albeit uncomfortable, finding. All other things equal, Muslims have faced barriers to economic integration in France that are higher than they would have been if everything about them were the same save for their religion. Part II of the paper explored the mechanisms through which these barriers were sustained. Here we find three channels through which our targeted group of Muslims faces higher barriers to economic success in France than does their matched group of Christians. Concerning tastes, we find no taste based discrimination by FdS against SM’s that is greater than towards SX’s; however SM’s reveal a distaste for the progress of SM-women, something that is not the case for SX’s towards SX women. Concerning beliefs, we find that FdS beliefs about SM’s drives distrust and discrimination. Also FdS and SM beliefs about other members of their own group limit generosity toward the other when more of their own group are observing. SM beliefs about the low possibility of cooperation among any groups with whom they interact in France, we infer from our data, accounts for their generalized low levels of altruism. Concerning technologies of communication, we find no evidence that SM are less able to communicate with FdS (and vice-versa). Overall, a range of mechanisms helps sustain religious inequality in France. The remedies for systematic inequality of opportunities are clearly a subject for future research. This paper’s contribution is to have identified that religious difference is a source of

21

inequality and specified some of the channels through which religious inequality flows.

22

TABLES

Table I-1. Impact of religion and geographic origin on household’s yearly income in large-n survey.

Variables Model (1) Model (2) coeff. s.e. coeff. s.e. Female -0.176** 0.041 -0.186** 0.042 Age -0.0272** 0.003 -0.030** 0.003 Education 0.227** 0.014 0.230** 0.017 Jewish 0.339* 0.165 0.499* 0.202 Asian religion -0.136 0.138 -0.396^ 0.208 Atheist -0.135 0.095 -0.082 0.106 Muslim -0.342** 0.047 -0.173 0.107 Number of years in France 0.008** 0.002 0.008** 0.002 Country Fixed Effects No Yes Observations 2,645 2,645 Pseudo R-squared 0.050 0.063

The table reports ordered probit estimates. The dependent variable is an ordinal variable ranging from the value “0” if the yearly household income is null to “14” if the yearly household income is greater than 68,000 Euros. Female is a binary variable, which takes the value “1” if the head of household is Female and “0” if the head of household is Male. Age is a continuous variable equal to the actual age of the head of household. Education is an ordinal variable ranging from the value “1” for no schooling to “6” for post-secondary education. Jewish is a binary variable, which takes the value “1” if the head of household is Jewish and “0” otherwise. Asian religion is a binary variable, which takes the value “1” if the head of household is Buddhist, Hindu, Shintoist or Confucianist, and “0” otherwise. Atheist is a binary variable, which takes the value “1” if the head of household is atheist and “0” otherwise. Muslim is a binary variable, which takes the value “1” if the head of household is Muslim and “0” otherwise. The reference group is Christian, which is a binary variable, which takes the value “1” if the head of household is Christian and “0” otherwise. Number of years in France is a continuous variable equal to the number of years that the head of household has been spending in France. Coefficients in bold highlight the effect of the confound between religion and country of origin. Standard errors are robust. Stars indicate coefficient significance levels (two-tailed): ** p<0.01, * p<0.05, ^ p<0.1.

23

Table I-2a: Results from the C.V. Experiment on Job Discrimination in France based on Religious Cues

Marie/Khadija Diouf Aurélie Ménard Difference Probability that candidates gets a call back: Marie/Aurélie matchup

0.21 (n=138)

0.27 (n=138)

-0.06

Probability that candidates gets a call back: Khadija/Aurélie matchup

0.08 (n=133)

0.25 (n=133)

-0.17**

Difference in difference significant at p<.01 Table I-2b: C.V. Results in Logit Regressions DV: likelihood that test candidate gets a call-back

(1) (2) (3)

Test candidate is Khadija -1.10** (0.38)

-1.10** (0.38)

-1.11** (0.39)

CDD contract 0.77* (0.38)

0.79* (0.38)

0.73^ (0.39)

Small Firm -0.09 (0.58)

-0.04 (0.57)

-0.12 (0.59)

Medium Firm 0.27 (0.62)

0.27 (0.62)

0.12 (0.62)

Photo included -0.25 (0.46)

-0.24 (0.47)

Secretary -0.53 (0.57)

Executive assistant -0.76 (0.53)

Junior accountant -0.03 (0.44)

Constant -1.77** (0.62)

-1.76** (0.62)

-1.39* (0.66)

Observations 265 265 265 Pseudo-R2 0.06 0.07 0.08 Logit regression of the likelihood that the test candidate, Khadija or Marie Diouf, gets a call-back. Omitted categories for variables on the right-hand side are as follows: Marie Diouf for the test candidate; CDI contract type; Large firm size; no photo included in the CV; Accountant for the job position. Standard errors are robust. Stars indicate significance: ^ p ≤ 0.10; * p ≤ 0.05; ** p ≤ 0.01.

24

Table I-3. Ordered probit estimates: Impact of religion on household’s monthly income. Variable Coefficient standard error Christian household 0.357** 0.127 Head of household’s gender 0.293* 0.120 Head of household’s education 0.055* 0.025 Education of the first migrant 0.022 0.035 Pseudo R2 0.017 Observations 312 The dependent variable is an ordinal variable ranging from the value “1” if the monthly household income is lower than 500 Euros to “9” if the monthly household income is greater than 7,500 Euros. Christian household is a binary variable, which takes the value “1” if the household is Christian and “0” if the household is Muslim. Head of household’s gender is a binary variable, which takes the value “1” if the head of household is Male and “0” if the head of household is Female. Head of household’s education controls for the level of education of the head of household. This is an ordinal variable ranging from the value “1” for no schooling to “8” for post-secondary education. The variable Education of first migrant controls for the level of education of the head of household’s ancestor who was the first to migrate to France, and thus absorbs the differences in current family income due to initial differences in human capital. This is an ordinal variable ranging from the value “1” for no schooling to “6” for post-secondary education. Results hold when we control for the subject’s ethnicity. Results hold as well in OLS regressions. Standard errors are robust. Stars indicate coefficient significance levels (two-tailed): ** p<0.01, * p<0.05, ^ p<0.1.

25

Table II-1: Games, Rules, and Pay-offs35

Game

Rules Maximum Returns to a player in Euros

Pre-test altruism game

1. Subjects sit in front of a computer, and view three screens sequentially, each screen with two pictures of confederates whose photos were taken in the district where the registration took place; 2. For each screen, subjects were given two Euros, and three envelopes. One envelope was addressed to the first face on the screen; one to the second; and one to the subject; 3. Subjects were told that they could not put two coins in any one envelope, but could put the two Euros in any two of the three envelopes; the money addressed to the faces would be donated to them; the money subjects put in their own envelope would stay with them to bring home; 4. Our measure of altruism was the amount (from 0-3) that the subject put in his/her three envelopes.

3

Trust 1. Subjects sat quietly in a waiting room (and supervised such that they could not communicate with one another), and were called to a playing table in pairs, but not knowing how many times they would be playing, nor against whom; 2. For each pair, one was assigned role of “sender” and the other “receiver” 3. Sender had 3 Euros in his/her account, and could send any amount {0, 1, 2, 3} to receiver by marking this amount on a sheet that receiver would never see; 4. All subjects were told before the game began that the amount sender sent to receiver would be tripled in value; they were also told the game was about “trust”, such that the more the sender sent to receiver, the greater total amount the second player would receive, and that the sender could do better by sending more money to the receiver if he/she trusted that receiver would return generously; 5. Receiver simultaneously marked on his/her sheet what percentage of the amount received would be sent back to sender [0, 1/3, 2/3, all]. 6. After each play, sender and receiver return to the waiting room, not knowing if they would be called again, nor in what role.

9

Speed Chatting 1. The 10 subjects for each experimental session were broken up into two teams of 5, each following the same protocol; 2. Each player on a team knew that he/she would have a few minutes to meet (and we emphasized, to get to know about) each other member, thereby “speed chatting” with four other players, sequentially, as in a speed-dating situation;

32

35 . All games had safeguards to meet IRB regulations; they are not part of the description here of the games.

26

3. After meeting each partner, players were given 1 minute to jot down notes on a piece of paper. 4. After meeting all other members of their group, each player received an answer sheet with the picture of each person they met, and a series of eight personal questions about them (e.g. their age, their job, whether they were married, their religion…). Players were allowed to consult their notes; 5. For each question subjects gave their answer [answer or “don’t know”], and circled whether they learned this information from their chat, or just guessed the answer; 6. For each correct answer for the 8 questions on 4 other players, subjects earned 1 Euro;

Friend and Job Recommendation

1. At the end of the questionnaire for each person they met in speed chatting, subjects were asked to rank on a scale from 1 to 10 whether (a) they would want to befriend this person and (b) whether they would recommend this person for a job.

0

Leader voting and distribution

1. After the speed chatting, each of the two teams were tasked with choosing a leader, one for each team; 2. Players knew that the leader would have the responsibility of dividing 30 Euros in any way he/she wanted (including keeping it all to her/himself) 3. Each player received a hand-out sheet with the pictures of each of the four other players on his/her team and ranked them as to whether they would want that person as a leader; 4. Players also indicated how much they would allocate to each of the players on the team (including themselves) were they to be elected the leader; 5. The player with the highest ranking in votes became the leader; his/her allocations were distributed.

30

Dictator 1. All 10 players sat together in a single room in front on a large screen; 2. They observed six distinct pictures of confederates in sequence on that screen. Four of the confederates were pictures of people solicited in the district where the experiment took place, two of them apparently FdS with typical FdS names; two of them apparently foreigners (“North African/Middle Eastern”). Finally, two confederates were a Senegalese man and a Senegalese woman; these people lived in other cities in France. These were our target confederates. 3. For half of the sessions, subjects viewed one of the neighborhood foreigners and one of the target confederates with a Christian name and the other with a Muslim name; for the other half of the sessions, this was reversed. 4. Subjects were given 5 Euros to distribute to each confederate; that which was not given was kept in the account of the subject to take home.

30

Strategic dictator 1. After the completion of the dictator allocation, one of the monitors explained that one of the players would be chosen at random as the “model”. This was a lie. In fact, the model was always one of the players who was known beforehand as an FdS.

30

27

2. Players were then told that they would see the same six photos that they saw in the dictator game. Now they would be asked to guess the amount the model allocated to each of the confederates in the dictator game; 3. The player who guessed most closely to the actual decisions of the model would receive a prize of 30 Euros.

Table II-2: Trust Game – Difference of Means Variable SM sender SX sender Difference H0

FdS return to SM and SX 0.39 (N=28)

0.51 (N=23) 0.11 Rejected

(p=0.09) Means tests were conducted using two-tailed t-tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the receiver faces a SM sender and for the scenario where the receiver faces a SX sender. When we run OLS regressions, which allow us to control for gender, age, education, income, whether the players know each other, whether the players know someone who participated in previous sessions, session fixed effects and to cluster standard errors at the player level, this result becomes significant at 95%.

Table II-3: Game Stakes and SM Altruism

FdS SX SM P value of t-tests Perceived maximum

stake Practice Altruism (amount given)

2.04 1.73 2.06 p=.17 (SMSX); p=.48 (SMFdS); p=.21 (SXFdS)

3

Trust Return .47 .44 .37

p=.11 (SMSX); p=.03 (SMFdS); p=.30 (SXFdS)

9

Leader Distribution (amount given per cap)

4.29 4.36 3.05 p<.001 (SMSX); p=<.001 (SMFdS);

p=.42 (SXFdS) 30

Dictator 1.97 1.74 0.79

p<.000 (SMSX); p<.001 (SMFdS); p=.12 (SXFdS)

30

28

Table II-4: In-group gender discrimination in the dictator game Average donation to

in-group male (max = 5 Euros)

Average donation to in-group female (max = 5 Euros)

Difference of means

SM 1.38 0.2 -1.18*** SM males 1.33 0.14 -1.19*** SM females 1.40 0.33 -1.07^ SX 1.86 2.5 0.64 SX males 2 3 N/A36

SX females

1.67 2.33 0.66

Table II-5A. Difference in the hopes for son’s vs. daughter’s educational level Variable SM

respondent SX

respondent Difference H0

Hopes of SM’s and SX’s for son’s vs daughter’s educational level

0.02 (N=302)

-0.22 (N=137) 0.24 Rejected

(p=0.03) The hopes for one’s child’s educational level is measured by a scale from 1 (“no diploma”) to 8 (“post-secondary education”). Regarding daughters, the average hope is 7.36 among SMs and 7.55 among SX’s. Regarding sons, the average hope is 7.38 among SM’s and 7.33 among SX’s. Means tests were conducted using two-tailed t tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the respondent is SM and for the scenario where the respondent is SX. When we run an OLS regression, which notably allows us to control for the gender, the age, the education level, the family income, and the time since arrival of the first migrant of the family, this result holds (significant at 90% with a p-value equal to 0.08).

Table II-5B. Difference in the perception of the purpose of women Variable SM

respondent SX

respondent Difference H0

Perception of the purpose of women by SMs and SXs

2.89 (N=358)

3.14 (N=146) -0.25 Rejected

(p=0.01) The perception of the purpose of women is measured by a scale from 1 (“completely agree”) to 4 (“completely disagree”) that captures the level of disagreement with the statement that the purpose of women is to have children and to raise them. Means tests were conducted using two-tailed t-tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the respondent is SM and for the scenario where the respondent is SX. When we run an OLS regression, which notably allows us to control for the gender, the age, the education level, the family income, and the time since arrival of the first migrant of the family, this result holds (significant at 99% with a p-value<0.01).

36 Difference-of-means test is not available here because our data rely on only one observation for SX male donation to SX female.

29

Table II-6A. Difference in the amount given by SM (resp. SX dictators) to FdS in sessions with a low number of SM (resp. SX) and in sessions with a high number of SM (resp.SX).

Variable

Sessions with low

number of SM

Sessions with high number of

SM

Difference

H0

Donation of SM to FdS 1.75 (N=4)

0.625 (N=32) 1.125 Rejected

(p=0.01)

Variable

Sessions with low

number of SX

Sessions with high number of

SX

Difference H0

Donation of SX to FdS 1.9 (N=10)

1.5 (N=12) 0.04

Failed to Reject

(p=0.12) Sessions with a low number of SMs (session 1 and session 7) comprise 10% of SM’s. Sessions with a high number of SMs encompass either 20% (sessions 2, 3, 4 and 6) or 30% (session 5 and session 8). Sessions with a low number of SXs (sessions 1, 3, 4, 5 and 7) comprise 10% of SXs. Sessions with a high number of SXs (sessions 2, 6 and 8) encompass 20% of SXs. Means tests were conducted using two-tailed t tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the number of SM/SX is low and for the scenario where the number of SM/SX is high. The difference in difference is significant at 99%. Table II-6B. Difference in the amount given by FdS dictators to SM confederates and to SX confederates in sessions with a low number of FdS and in sessions with a high number of FdS.

Variable

Sessions with low

number of FdS

Sessions with high number of

FdS

Difference

H0

Donation of FdS to SM 2.82 (N=11)

2.00 (N=12) 0.82 Rejected

(p=0.10)

Variable

Sessions with low

number of FdS

Sessions with high number of

FdS

Difference H0

Donation of FdS to SX 2.18 (N=11)

2.17 (N=12) 0.01

Failed to Reject

(p=0.49)

30

Sessions with a low number of FdS (sessions 2, 3, 6, 7 and 8) comprise 20% of FdS. Sessions with a high number of FdS’s comprise either 30% (session 4) or 40% (session 1 and session 5) of FdS. Sessions with a high number of SX’s (sessions 2, 6 and 8) comprise 20% of SX’s. Means tests were conducted using two-tailed t-tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the number of FdS’s is low and for the scenario where the number of FdS is high. The difference in difference is significant at 99%. Table II-7. Difference in the belief of FdS about other FdS’ donation to SM and to SX

Variable

SM

SX

Difference

H0

Donation of the old FdS to SM/SX 0.17 (N=6)

0.63 (N=8) 0.46 Rejected

(p=0.08)

Variable SM

SX

Difference

H0

Donation of A3 to SM/SX 2.54 (N=13)

2.38 (N=13) -0.16

Failed to reject

(p=0.38)

Variable SM

SX

Difference

H0

Donation of the young FdS to SM/SX 1.4 (N=10)

0.7 (N=10) -0.70 Rejected

(p=0.07) Means tests were conducted using two-tailed t- tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the confederate is SM and for the scenario where the confederate is SX. When we run an OLS regression, which notably allows us to control for the gender, the age, the education level, and the family income of the FdS guessing about other FdS’s donation, these results hold. Table II-8A. Difference in the amount given by SM dictators (resp. SX dictators) to FdS confederates in sessions with a low number of FdS and in sessions with a high number of FdS.

Variable

Sessions with low

number of FdS

Sessions with high number of

FdS

Difference

H0

Donation of SM to FdS 1 (N=22)

0.36 (N=14) 0.64 Rejected

(p=0.03)

Variable

Sessions with low

number of FdS

Sessions with high number of

FdS

Difference H0

Donation of SX to FdS 1.75 (N=16)

1.5 (N=6) 0.25 Failed to

Reject

31

(p=0.26) Sessions with a low number of FdS (sessions 2, 3, 6, 7 and 8) comprise 20% of FdS. Sessions with a high number of FdS comprise either 30% (session 4) or 40% (session 1 and session 5) of FdS. Sessions with a high number of SXs (sessions 2, 6 and 8) comprise 20% of SXs. Means tests were conducted using two-tailed t-tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the number of FdS is low and for the scenario where the number of FdS is high. The difference in difference is significant at 99%.

Table II-8B. Difference in the amount given by FdS dictators to SM confederates (resp. SX confederates) in sessions with a low number of SM (resp. SX) and in sessions with a high number of SM (resp. SX). Variable

Sessions with low

number of SM

Sessions with high number of

SM

Difference

H0

Donation of FdS to SM 2.83 (N=6)

2.24 (N=17) 0.59

Failed to reject

(p=0.21)

Variable

Sessions with low

number of SX

Sessions with high number of

SX

Difference H0

Donation of FdS to SX 1.94 (N=22)

2.71 (N=14) -0.77 Rejected

(p=0.09) Sessions with a low number of SMs (session 1 and session 7) comprise 10% of SMs. Sessions with a high number of SM’s comprise either 20% (sessions 2, 3, 4 and 6) or 30% (session 5 and session 8). Sessions with a low number of SX’s (sessions 1, 3, 4, 5 and 7) comprise 10% of SX’s. Sessions with a high number of SXs (sessions 2, 6 and 8) comprise 20% of SX’s. Means tests were conducted using two-tailed t tests with paired variances for the two scenarios. For H0, means are equal for the scenario where the number of SM/SX is low and for the scenario where the number of SM/SX is high. The difference in difference is significant at 99%.

Table II-9: Speed Chatting Quiz Results

FdS about SX FdS about SM Difference of means Average # correct 3.53 out of 8 4.55 out of 8 -1.02*

% correct that was learned 67.02 60.44 -6.58

% correct that was guessed 25.83 35.05 9.22

SX about FdS SM about FdS Difference of means

Average # correct 4 out of 8 4.5 out of 8 -0.5 % correct that was learned 35.33 34.02 -1.31

% correct that 33.89 45.52 11.63

32

was guessed

33

Table III-1 Given Names and their Foreign-ness in a French Setting

[Names in Bold are “Recognizable” as Possibly French; Names in Italics are Common Muslim Names]

MUSLIM NAMES CHRISTIAN NAMES

AMADOU CECILE AMIE CHRISTINE

ASTOU DANIEL AWA EPHIGENIE CIRE GASTON

FATOUMATA HELENE KALS LOUIS

KHADY MAMADOU JEAN MAMADOU LAMINE NINA

MOUSTAPHA ROBERT ANTOINE NDEYE THERESE

OUSMANE SIDY SIRE

TAMSIR YACINE

34

References

Amadieu Jean-François (2004), Enquête « Testing » sur CV, Adia/ Paris I- Observatoire des discriminations. Bisin, Alberto, Eleonora Patacchini, Thierry Verdier, and Yves Zenou (2008), « Are Muslim immigrants different in terms of cultural integration? », Journal of the European Economic Association, 6, 445-456. Bowen, John (2009) Can Islam be French? Pluralism and Pragmatism in a Secularist State (Princeton: Princeton University Press) Caldwell, Christopher (2009) Reflections on the Revolution in Europe: Immigration, Islam, and the West (New York: Doubleday) Constant, Amélie, Liliya Gataullina, Klaus Zimmermann and Laura Zimmermann (2006) Clash of Cultures: Muslims and Christians in the Ethnosizing Process, IZA Discussion Paper Series 2350. Diop, A. Moustapha “Stéréotypes et stratégies dans la Communauté Musulmane de France”, pp. 77-98, in Leveau, Rémy and Gilles Kepel, eds. (1988) Les Musulmans dans la Société Française (Paris: Presses de la Fondation Nationale des Sciences-Politiques) Donselaar, Jaap van and Peter Rodrigues (December 2004), “ Monitoring Racism and the Extreme Right” Anne Frank Stichting, Research and Documentation, Leiden University, Amsterdam, Department of Public Administration (Fifth and Sixth Reports ). Duguet, Emmanuel, Noam Leandri, Yannick L'Horty and Pascale Petit (2008) Are Young French Jobseekers of Ethnic Immigrant Origin Discriminated Against? A Controlled Experiment in the Paris Area, Working Paper. Jowell, R. and Prescott-Clarke, P. (1970). Racial Discrimination and White-collar Workers in Britain. Race and Class, 11; 397. Kepel, Gilles (1987) Les Banlieues de l’Islam (Paris: Seuil) Laurence, Jonathan and Justin Vaisse (2006) Integrating Islam: Political and Religious Challenges in Contemporary France (Washington, DC: Brookings) Manning, A. and S. Roy (2007), “Culture Clash or Culture Club? The Identity and Attitudes of Immigrants in Britain.” CEP Discussion Paper No. 790, London School of Economics. Meurs, Dominique, Ariane Pailhé, and Patrick Simon “the Persistence of Intergenerational Inequalities linked to Immigration: Labour Market Outcomes for Immigrants and their Descendants in France” Population (E 2006): 645-82

35

Norris, Pippa and Ronald Inglehart (2003) “Islamic Culture and Democracy: Testing the ‘Clash of Civilizations’ Thesis”. In Ronald Inglehart, ed. Human Values and Social Change: Findings from the Values Surveys (Leiden: Brill), pp. 5-32

Wilson, Rick and Catherine Eckel, “Judging a Book by its Cover: Beauty and Expectations in the Trust Game” Political Research Quarterly, Vol. 59, No. 2, 189-202 (2006) Wolff, François-Charles, Seymour Spilerman, and Claudine Attias-Donfut (2007) “Transfers from Migrants to their Children: Evidence that Altruism and Cultural Factors Matter” Review of Income and Wealth 53(4): 619-644

36

Appendix

Table A-1. Descriptive statistics in the Laitin CSA Survey Variable Observations Mean S.D. Min. Max. Household monthly income

461 5.15 1.66 1 9 312 5.10 1.64 1 9

Christian household 509 0.29 0.45 0 1 312 0.32 0.47 0 1

Head of household’s gender

439 0.67 0.47 0 1 312 0.65 0.48 0 1

Head of household’s education

400 4.63 2.63 1 8 312 4.74 2.68 1 8

Education of the first migrant

397 3.00 1.82 1 6 312 3.02 1.85 1 6

For each variable, the first row presents descriptive statistics over the whole sample (N=511) while the second row presents descriptive statistics over the regression sample used in Table I-3 (n=312). Table A-2: Balance tests between Senegalese Muslims and Senegalese Christians in Game Experiments SM SX Difference of means

(two-tailed tests assuming unequal variances)

% Female 0.44 0.55 -0.10 Age 33.56 31.45 2.10 Education 7.59 7.63 -0.04 % Self-employed 0 0.14 -0.14 Religiosity 2.71 4.9 -2.19** Ideology 4.73 4.43 0.30 Distrust 2.79 2.33 0.45^ % Born in France 0.06 0.09 -0.04 Play for money 5.56 4.22 1.34 Family income 3.87 4 -0.13 % Joola 0.61 0.5 0.11 % Serer 0.33 0.2 0.13 % French national 0.17 0.18 -0.02 Job skill level 2.8 2.29 0.51 % Reside in Paris 0.53 0.7 -0.17 Education is an ordinal variable ranging from the value “1” for no education to “9” for post-secondary education ; Religiosity is an ordinal variable ranging from the value “1” if the player never attends religious services to “7” if the player attends religious services several times per week; Ideology is thermometer measure ranging from “1” for left-most political ideology to “10”

37

for right-most ideology; Distrust is an ordinal variable ranging from the value “1” if the player believes one can almost always trust people to the value “4” if the player believes one can never be careful enough when dealing with others; Play for money is a thermometer measure ranging from “1” for least likely to have played only for money to “10” for most likely to have played only for money; Family income is an ordinal variable ranging from the value “1” for a household monthly income below 500 Euros to “11” for a household monthly income above 7,500 Euros; Job skill level is an ordinal variable ranging from the value “1” for high-skill to the value “5” for unemployed.

38

Appendix: Examples of the CVs

Marie DIOUF 14 rue Gérard 75 013 Paris Tel : 06 46 59 74 92 e-mail : diouf_marie@voila.fr 24 ans, célibataire Nationalité française, Permis B

ASSISTANTE POLYVALENTE

Domaines de compétences

Accueil physique et téléphonique Tenue d’agenda Secrétariat classique Traitement du courrier : rédaction, frappe et envoi des lettres, offres de prix, contrats Prise de rendez-vous et organisation des réunions et des séminaires Organisation des déplacements : demande de visa, réservations d’avion Suivi des frais généraux Préparation et saisie des budgets de ventes Suivi des budgets Maîtrise de l’anglais et l’allemand professionnels

Formation

2005 BTS Assistante Secrétaire Trilingue (Anglais, Allemand, Français)

Expérience Professionnelle

39

Nov. 2007 – Mars 2009

Août 2006 – Oct. 2007

Oct. 2005 – Juillet 2006

Avril 2005 – Sept. 2005

Secrétaire polyvalente NOVOTEL Paris :

- Rédaction et frappe de courriers - Mise à jour du registre unique du personnel - Contact avec les administrations (sécurité sociale…) - Demande d’affiliation complémentaire santé - Secrétariat classique

Secrétaire SCP GAUTIER, VROOM & ASSOCIES à Saint Maur de Fossés :

- Frappe de courriers et actes divers à l’aide du dictaphone - Accueil téléphonique de la clientèle - Relances factures clients - Prise de rendez-vous - Mise sous pli - Affranchissement du courrier - Classement - Télécopie - Reliure de documents

Assistante polyvalente Secours Catholique à Evry :

- Traitement des appels téléphoniques - Organisation de réunions, colloques et séminaires - Accueil téléphonique et physique - Rédaction, mise en forme de documents, classement et archivage - Prise de notes et réalisation de comptes rendus

Assistante polyvalente Secours Catholique à Evry :

- Gestion de planning - Accueil téléphonique et physique - Gestion de dossiers - Organisation de réunions

Autres activités

Responsable bénévole aux Scouts et Guides de France

40

Aurélie Ménard Célibataire (03/07/1985) 45 Rue Tolbiac aurel.menard@yahoo.fr 75013 PARIS 06 48 20 74 08

Assistante polyvalente (disponibilité immédiate)

Compétences

Accueil physique Accueil téléphonique (Filtrage d’appels ; mise en relation avec le service adéquat) Organisation de réunions Tenue des agendas et prise de rendez-vous Plannings : réunions, salles de réunion, congés, déplacements Rédaction et mise en forme de documents, envoi de mailings, rédaction de compte

rendu Gestion du courrier postal et électronique (Réception ; Emargement ; Diffusion ;

Mailing ; Envoi de documents commerciaux ; Facturation) Commandes fournisseurs, commandes de fournitures, gestion du parc téléphonique Préparation et coordination d’évènements spéciaux (salons commerciaux, portes

ouvertes, lancement de nouveaux produits) Classement ; Archivages Informatique: Word/Excel/PowerPoint/Office Outlook/Internet Langues: Anglais - Allemand

Expériences professionnelles 2007 - 2009 Assistante polyvalente CODEM TECHNOLOGIES

(Fabrication et vente de produits électroniques de sécurité)

2007 Assistante polyvalente ESTEL (Etudes de marché) 2005 - 2007 Secrétaire ORGECO

41

(Fabrication et vente de bijoux fantaisie)

Formation :

2005 Brevet de Technicien Supérieur – Assistante de Direction Stage de fin d’études de 4 mois à la CAF de Lille 2003 Baccalauréat Professionnel Bureautique Secrétariat 2001 Brevet d’Etudes Professionnelles Bureautique Secrétariat

Loisirs : - Volley-ball (6 ans) - Lecture, cinéma - Permis de conduire [it is strange for this informations to be mentioned under the “Loisirs”

heading…]

42

Khadija DIOUF 22 rue de Cotte 75012 PARIS Tel : 06 48 20 28 29 e-mail : khadijadiouf@ymail.com 24 ans, célibataire Nationalité française, Permis B

ASSISTANTE POLYVALENTE

Domaines de compétences

Accueil physique et téléphonique Tenue d’agenda Secrétariat classique Traitement du courrier : rédaction, frappe et envoi des lettres, offres de prix, contrats Prise de rendez-vous et organisation des réunions et des séminaires Organisation des déplacements : demande de visa, réservations d’avion Suivi des frais généraux Préparation et saisie des budgets de ventes Suivi des budgets Maîtrise de l’anglais et l’allemand professionnels

Formation

2005 BTS Assistante Secrétaire Trilingue (Anglais, Allemand, Français)

Expérience Professionnelle

43

Nov. 2007 – Mars 2009

Août 2006 – Oct. 2007

Oct. 2005 – Juillet 2006

Avril 2005 – Sept. 2005

Secrétaire polyvalente NOVOTEL Paris :

- Rédaction et frappe de courriers - Mise à jour du registre unique du personnel - Contact avec les administrations (sécurité sociale…) - Demande d’affiliation complémentaire santé - Secrétariat classique

Secrétaire SCP GAUTIER, VROOM & ASSOCIES à Saint Maur des Fossés :

- Frappe de courriers et actes divers à l’aide du dictaphone - Accueil téléphonique de la clientèle - Relances factures clients - Prise de rendez-vous - Mise sous pli - Affranchissement du courrier - Classement - Télécopie - Reliure de documents

Assistante polyvalente Secours Islamique à Massy :

- Traitement des appels téléphoniques - Organisation de réunions, colloques et séminaires - Accueil téléphonique et physique - Rédaction, mise en forme de documents, classement et archivage - Prise de notes et réalisation de comptes rendus

Assistante polyvalente Secours Islamique à Massy :

- Gestion de planning - Accueil téléphonique et physique - Gestion de dossiers - Organisation de réunions

Autres activités Responsable bénévole aux Scouts Musulmans de France