CAIRN-INT.INFO : International Edition

1There has been relatively little sociological work in France on legal responses to racist speech and acts,  [1] in spite of the existence of philosophical and legal research into French and international hate speech law,  [2] parliamentary debates over the law of July 1, 1972, on racist speech,  [3] court proceedings,  [4] comparisons between legislation in the United States and Europe,  [5] and the issue of double standards in the legal treatment of Islamophobic speech.  [6] The only sociological study, conducted by Erik Bleich, is lacking both theoretically (the concept of an “enforcement rate” is unsatisfactory)  [7] and methodologically (the study is based on just 103 judgments by the Court of Cassation, between 1972 and 2012). As Bleich himself acknowledges, studies of the courts’ treatment of hate speech ideally require “the ability to review data on complaints, prosecutions, convictions, and sentences related to hate speech.”  [8] Furthermore, there has been little in-depth sociological research in France on legal responses to hate crimes, and particularly to racist acts. Following a rise in anti-Semitism in the early 2000s, one research team did examine the “lure of anti-Semitism.”  [9] However, the available data are based primarily on (incomplete) statistics from the ministry of the interior, and it is difficult to generalize from the few cases where the suspect’s profile is known. Furthermore, the study focuses on anti-Semitism to the exclusion of all other forms of racism.

2The gaps in French approaches to racist offenses do not extend to other sorts of crime. There is a long tradition in the sociology of justice that recognizes that various social factors influence the legal process.  [10] Focusing exclusively on non-racist offenses that resemble hate crimes, including violence against women,  [11] a number of studies have highlighted the social factors that influence the approach of the police and courts, particularly in cases of sexual violence and/or marital rape.  [12] Our work continues this research tradition, while building on the foundations laid by hate crime studies.

3The lack of French research into racist offenses contrasts with the high number of US and, to a lesser extent, British studies. These have revealed several factors that determine how the police prosecute cases, including the legal definition of hate crimes,  [13] the location of police services,  [14] the existence of specialized hate crimes units, whether the legal classification of the offense is subject to review,  [15] the type of offense, the type of prejudice involved, the relationship between victim and suspect,  [16] and the suspect’s ethnicity, with racial minorities being, paradoxically, over-represented among suspects.  [17]

4While the police work itself has been studied closely, the early stages of the process are less well understood, and little is known about hate crime training in police academies beyond the content of the textbooks used.  [18] In particular, there is little information about the role of the prosecutor’s office, the equivalent of the city or district attorney or attorney general in the United States. As a 2012 study points out, “very little attention has been given to the topic of hate, or bias crime, prosecution.”  [19] Research has nonetheless shown that such cases are highly unlikely to be prosecuted or result in a conviction and has explored the reasons for this. There are, first of all, legal reasons: the supreme court’s more or less expansive definition of hate crime;  [20] the inclusion of race, ethnicity, religion, and sexual orientation as defining criteria for protected groups in state hate crime laws, and special provisions for vandalism;  [21] and the difficulty of proving an incident was motivated by hate,  [22] which leads police and prosecutors to select only the most “winnable” cases.  [23]

5Next, there are professional reasons. A study of professional practice has shown that decision-making processes in hate crime cases broadly resemble those in other types of case.  [24] As a result, the prosecution rate is similar to cases that do not involve prejudice, and even higher in cases involving religion, or where there are multiple motives.  [25] However, it has also been shown that hate crimes are less likely to be prosecuted if they are viewed as “low-level”  [26] or less serious offenses,  [27] if the perpetrator cannot be located or is unknown to the victim,  [28] or if prosecutors are unfamiliar with hate crime law.  [29] Conversely, such cases are more likely to be taken up if the prosecutor’s office has an active antiracism policy,  [30] dedicated “community relations” staff,  [31] or ongoing relationships with antiracist organizations.  [32]

6Finally, there are extralegal factors. These may include political pressure: public prosecutors are elected, and prosecuting hate crimes may be viewed as a political means to win support from ethnic minorities.  [33] One study has identified other factors influencing prosecutors’ decisions, including higher proportions of same-sex households and low levels of “Protestant fundamentalism” in a jurisdiction.  [34]

7Nevertheless, most research has focused on legal factors. Professional and extralegal factors have not been examined systematically and, paradoxically, other factors related to the case itself have simply been ignored. Our hypothesis is that the judicial process is shaped by interactions between a number of factors: the characteristics of the cases and actors involved, and professional, legal, and extralegal factors. Our work lies at the intersection of US sociological research into hate crimes—although we examine the work of prosecutors as well as police—and French approaches to the sociology of justice, which examine how the treatment of various offenses by the police and courts reproduces inequality.  [35] The hypothesis that a range of factors interact in such cases allows us to identify anything potentially distinctive about the way the legal system treats racist offenses in comparison to others. Consequently, we will examine the work of the public prosecutor’s office and the phenomenon of attrition during the legal process, from the initial complaint to sentencing. Our second hypothesis concerns the possible effects of the social characteristics of those involved in racist offenses: does the interaction of social class, gender, ethnicity, and age have a differential effect on the treatment of complainants and suspects? In this respect, our study follows in the wake of the intersectional sociology of law.  [36]

8Our investigation examines these two hypotheses from the constructivist “everyday racism” approach,  [37] which assumes that an event becomes “racist” through the action of social actors. The label “racist” is the product of interactions and negotiations between actors, following logics that may converge or diverge. So, we must go behind the scenes of police and legal work and study how legal classifications are made  [38]—something that illustrates the discretionary power administrative agents enjoy.  [39] We pay particular attention to commonsense reasoning and reasoning practices in law enforcement.  [40] More specifically, we analyze how racist offenses can be understood through administrative logics (categories of administrative reasoning) and commonsense social logics (categories of social reasoning).  [41]

9Our aim is to examine the factors that determine decisions by public prosecutors, using both quantitative data (500 coded cases from three criminal courts) and qualitative data (12 interviews with prosecutors and deputy prosecutors, including several prosecutors with specific responsibility for racist offenses). We can understand the phenomenon of attrition—the way cases are filtered out during the legal process depending on certain characteristics—by cross tabulating cases. Meanwhile, logistic regression provides information about the probability, all other things being equal, that an offense will be verified (the “[un]verified racist offense” variable)  [42] depending on a series of variables (location, type of offense, type of racism, profile of actors involved, etc.).

10To explain how the legal system treats racist offenses, we show that, in spite of significant legislative and institutional differences between France and the United States, the two countries share similar social and professional logics. Various things influence the way these offenses are dealt with: legal and political factors, organizational and practical factors, the professional definition of what constitutes a racist incident, the type of racism involved, and the profiles of the suspect and the victim. This combination of factors typically lends priority to the more “serious” offenses, and to forms of racism that can be legitimately combated. Finally, such treatment has the paradoxical effect that racial minorities are over-represented among offenders, in spite of the fact that antiracism legislation is supposed to protect them.

Criminal policy and legal technicalities

11Public prosecutors in France have greater power to direct police work and to decide how to approach a case than in the United States or the United Kingdom. They direct the police investigation, rather than having to wait for it to conclude. Nonetheless, the prosecutor’s assessment depends on the victim’s perception of the crime and the police’s interpretation of it:


We have to assume that [when] discrimination cases are entrusted to a prosecutor, once the complaint has been taken down and recorded [electronically] it has to be highlighted. If it’s just described as violence but without mentioning the origin [of the victim], etc., it gets handed to the prosecutor who deals with violence. (Interview with a deputy public prosecutor, racism specialist)

13Unfortunately, we were unable to study this important moment in the fight against racism, as the police authorities refused to allow interviews. As one prosecutor explains: “It happens at the level of the police, but normally we oversee it. These cases are reported to the [prosecutor] hotline, so it’s up to them to ask the right questions, to find out what area we’re in, and to correct the [legal] classification” (interview with a prosecutor). While police officers have some autonomy from prosecutors, the latter have substantial discretion in deciding whether to classify a case as a racist offense.

14This discretionary power is influenced primarily by legal and political factors: the prosecution-centered definition of legal classification, the highly technical nature of press law, and the priorities of criminal policy. Indeed, “how much hate crime there is and what the appropriate response should be depends upon how hate crime is conceptualized and defined.”  [43] The perception of the victim is central to the expanded definition of hate crimes used in London, whereas New York uses a much more restricted definition, based primarily on the police’s discretion.  [44] In France, the initial legal classification at the time of the complaint is decided jointly by the complainant and the police officer, and it depends on the victim’s sense of the racist nature of the offense and on how well the police officer has been trained in antiracism law. Still, the public prosecutor’s office has the final word. The initial legal classification is not victim-centered, and only partly police-centered. Above all, it is prosecution-centered, and involves a distinctive problem in French law: while racist acts are punishable under criminal law, racist speech falls under press law (Law of July 29, 1881, revised by the Law of July 1, 1972). This highly technical branch of law, which aims to protect freedom of speech, creates a major legal barrier to punishing racist discourse: “The only drawback of this legislation is that it falls under press law. [...] It’s complicated, legally. [...] [If it were] a normal offense, to be treated normally, things would be simpler” (interview with a prosecutor). Prosecutors are constrained by a short limitation period (three months for nonpublic speech, and a year for public speech), and must be careful about the “publicity” of the insult—the border between “private” and “public” being the subject of complex legal argument.

15As well as these legal constraints, there are also political constraints, i.e., those created by both local and national criminal policy. Each court sets its own priorities, depending on the character of crime in the region. In court 1 (C1), for example, “priority is given to offenses against persons, particularly domestic violence. Next [come] property and traffic offenses. Those are the three big areas we primarily work on. Everything else is on the periphery” (interview with a prosecutor).

16Racist offenses are among the “peripheral” offenses, and only a few figure among the mass of cases handled by the prosecutor’s office. For example, in 2015, court 3 (C3) dealt with 115,335 cases, only 208 (0.18%) of which involved racism, while court 2 (C2) dealt with 49,023 cases, and only 57 cases involving racism (0.12%).

Table 1. Regression analysis results
Actors_Accused_SocialClass.Middle (ref.)ref.ref.ref.ref.ref.ref.ref.
Actors_Accused_Conviction.None (ref.)ref.ref.ref.ref.ref.ref.ref.
Actors_Accused_Conviction.More than one conviction13.381.12.350.0191.54115.9*
Actors_Accused_Conviction.One conviction4.420.91.650.10.7625.73.
Actors_Accused_Sex.Female (ref.)ref.ref.ref.ref.ref.ref.ref.
Actors_VictimReaction.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Actors_Relation.Administrative (ref.)ref.ref.ref.ref.ref.ref.ref.
Actors_Relation.No relation1.850.860.720.4730.359.94.
Actors_Relation.Distant acquaintance6.591.451.30.1940.39112.88.
Actors_Subordination.Accused > Victim0.590.64-0.820.4110.172.06.
Actors_Subordination.No subordination relationship (ref.)ref.ref.ref.ref.ref.ref.ref.
Actors_Subordination.Victim > Accused4.030.642.170.0311.1514.15*
Actors_Victim_SocialClass.Middle (ref.)ref.ref.ref.ref.ref.ref.ref.
Actors_Victim_Sex.Female (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_Xenophobic.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_Antisemitic.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_AntiNorth African.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_AntiNorth African.Yes3.170.681.70.0910.8412.04.
Incident_AntiMuslim.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_AntiBlack.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_Vandalism.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_AntiWhiteHostility.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_Abuse.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_Provocation.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Incident_Violence.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Procedure_Court.C1 (ref.)ref.ref.ref.ref.ref.ref.ref.
Procedure_PreviousConflict.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Procedure_CrossComplaint.No (ref.)ref.ref.ref.ref.ref.ref.ref.
Procedure_ResponseCrime.Complete denial (ref.)ref.ref.ref.ref.ref.ref.ref.
Procedure_ResponseCrime.Partial admission4.290.463.160.0021.7410.58**
Procedure_ResponseCrime.Complete admission4.830.722.20.0291.1919.62*
Procedure_SubstituteNumber.One (ref.)ref.ref.ref.ref.ref.ref.ref.
Procedure_SubstituteReference.No (ref.)ref.ref.ref.ref.ref.ref.ref.

Table 1. Regression analysis results

. p ≥ 0.05; * p < 0.05; ** p < 0.01; *** p < 0.001
Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017.
Population: Cases in which victims have made a complaint, involving a victim and a suspect who are physical persons, excluding unfounded cases.
N = 252, weighted data.

Limited material resources, administrative routines, and the hierarchy of crimes

17The prosecutor’s discretionary power is also determined by organizational and practical factors: the material constraints of the court, the amount of work to be done, the type of offense, the prosecutors’ degree of specialization, and how prosecutors anticipate that the courts will treat the case. The court is limited by its capacity to handle cases: “I have 40,000 proceedings coming in [each year, and] 10,000 that I think are prosecutable. Among those, I have the capacity to organize about 3,000 court hearings. [...] There’s a physical limit. [...] for less serious offenses, 2,000 cases is the absolute maximum” (interview with a prosecutor). These material constraints lead to a preference for the most “serious” cases, which “deserve” to be tried: “This is really one of the areas [i.e., violence] where there’s zero tolerance about prosecution. [...] With [violent] racist acts, we automatically go to a hearing” (interview with a prosecutor).

18Logistic regression shows that the factor that contributes the most to unverified racist offenses  [45] is being located in C3 (see Table 1).  [46] This local difference is probably due to organizational factors: C3 handles far more cases than C1. Such pressure on the legal system is a consequence of the mismatch between large quantities of cases and limited resources, and leads to the development of a hierarchy of offenses that deserve prosecution.  [47] The greater the pressure, the stricter the hierarchy of offenses, and the greater the number of offenses viewed as “peripheral” and not prosecuted.

19Indeed, another determining factor is the type of offense. Sorting cases by type of offense reveals the hierarchy of crimes (see Table 2). The proportion of suspects whose cases were unverified provides a good illustration of this hierarchy.  [48] The most regularly punished offense is contempt. These are instances of verbal abuse and/or resistance toward authority figures (e.g. police officers) and public servants (e.g. teachers). The absolute number of such cases is low (n=15), and we include them because, in addition to verbal abuse, the accused made statements that were considered racist. Contempt cases represent only 3.3% of all suspects, but none were not prosecuted, 93.3% were prosecuted,  [49] 86.7% went to court, and 46.7% ended in a guilty verdict. This unsurprising result is explained by the existence of clear evidence, by the fact that the victims represented public authority, and by prosecutors’ and judges’ intolerance of such crimes.

Table 2. Cases by type of offense

OffenseNot availableOther grounds for non-prosecutionNon-prosecution: Unverified racist offenseCase prosecutedIncluding case going to courtIncluding guilty verdictTotal suspects
Verbal abusen.131031631206139399

Table 2. Cases by type of offense

Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017. Calculations based on the total number of cases with at least one item of data available on the suspect. Interpretation: Among the 458 suspects (physical persons) in these cases, 18.8% appeared at hearings and 9.6% were found guilty. 46.7% of suspects in contempt cases were found guilty, whereas none were in cases of vandalism and only one in cases of discrimination and defamation.

20Moreover, racist violence is considered more “serious” than verbal abuse, and prosecutors are quick to prosecute such cases if there is sufficient evidence,  [50] as is often true in cases of violence (for instance, a medical certificate recommending that the victim temporarily stay home from work). Moreover, unlike cases of incitement to racial hatred and vandalism, discrimination cases are more often not prosecuted.  [51] Such offenses are less likely to be punished because it is very difficult to prove discrimination, especially when the statements gathered by the police are contradictory. In any case, discrimination appears in only twenty-nine cases (5.8%) in our sample, and is one of the least reported offenses.

21Another organizational factor is the prosecutor’s level of specialization. Legally, each court must have a deputy prosecutor in charge of racist offenses.  [52] Cases handled by deputy prosecutors are 3.03 times more likely to be unverified than those handled by others (see Table 1). This contradicts several US surveys, which found that the more specialized a police officer in the area of hate crime, the more likely he or she is to agree with the initial classification of the offense as racist. Conversely, leaving officers to their own devices—without training or legal classification review—makes it more likely that they will adopt a restrictive definition of racism.  [53] This surprising result shows that practices vary greatly, and that a deputy prosecutor in charge of racism cases is far from being specialized. Indeed, there are certainly individuals in charge of racist offenses among the deputy prosecutors in C1 and C2, but they have not received any initial training on racism by the National School of Magistrates, or any continuing education. As in any other area of law, they get on-the-job training by reading justice ministry circulars. Since there are a low number of racism cases, these deputy prosecutors are simultaneously in charge of other areas of law—for instance, road traffic offenses. Strictly speaking, they are not specialists in racism cases, as they do not devote the majority of their time to such cases. Deputy prosecutors tend to treat racism cases like any other. Moreover, it is rare for them to discuss cases with other magistrates. Unlike some US police departments, there is no further assessment of racism cases. The prosecutor alone makes the decision.

22Another criterion for prioritizing offenses is the anticipation of judgments, as prosecutors prefer to take up a case when they are relatively sure of a conviction. This makes them reluctant to prosecute cases where there is any doubt that the suspect was motivated by racism:  [54]


If you’re really dealing with classic violence, where there’s no particularly obvious racist motive, you’re going to drop [the aggravating circumstance], because you have to keep in mind that, if you decide about a classification and go to court, you’re going to have to defend it and you’re going to have to convince people. If you’re not convinced yourself, it can be very tricky. (Interview with a prosecutor in charge of racism cases)

24Thus, the public prosecutor’s work is shaped by the construction of a hierarchy of offenses based on their seriousness and the anticipated judgment: only those considered serious and “winnable” are sent to court. This judicial hierarchy of offenses resembles the hierarchy of police relevance observed in the United States and the United Kingdom,  [55] with its distinction between “good crimes” and “rubbish crimes.”  [56]

The professional definition of the ideal racist situation

25Furthermore, prosecutors’ discretion is influenced by administrative and social reasoning, which crystallizes in their professional definition of racist situations: treating racism cases as they would others, recognizing witnesses, defining an ideal type for racist situations, assessing the relationship between the victim and the suspect (whether a preexisting conflict exists, whether the victim “provoked” the suspect, whether the victim arouses suspicion, etc.), and defining the “good victim.”

26In reasoning about evidence and racist motivations, prosecutors treat racism cases as they do any other: “Even if they’re specific offenses, we approach them as we do any other” (interview with a deputy prosecutor in charge of racist offenses). Their reasoning resembles that of nonspecialized US police officers.  [57] Racist motivation “has to be deduced from verbalizations. [If] you attack a guy and beat him up without saying anything, we can’t call it a racist act” (interview with a prosecutor). Even if the victim interprets the violence as racist, the prosecutor will not follow it up without “verbalization.” However, the prosecutor may recognize such a motivation if certain “external elements” are present—for instance, if the accused is “dressed like a skinhead,” with “big combat boots” or “swastika tattoos.” Prosecutors’ evidential approach is typical of legal reasoning, where some pieces of evidence weigh more heavily than others.

27Although all racist abuse is by definition verbalized, in order to be considered credible the victim’s account must be verified by material evidence (written material, graffiti, etc.) or by external testimony:


This is the problem of proof: If I tell you something and there are only two of us, it’s one person’s word against another’s, which means that in some cases, if there’s nothing beyond that, we’ll have to drop the case, as we don’t have enough charges to prosecute. (Interview with a prosecutor)

29This is confirmed by the data.  [58] Situations that pit one person’s word against another’s are typically dismissed as “insufficiently characterized offenses”—particularly when the offense is a “less serious” one like racist abuse. It is no coincidence, then, that testimony in favor of the suspect can also be decisive.  [59]

Table 3. Cases with and without witnesses

WitnessesNot availableOther grounds for non-prosecutionNon-prosecution: Unverified racist offenseCase prosecutedIncluding case going to courtIncluding guilty verdictTotal suspects
For the victimn.2122351311888
For the accusedn.31424116152
No witnessn.78195633222246

Table 3. Cases with and without witnesses

Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017. Calculations based on the total number of cases with at least one item of data available on the suspect. Interpretation: 20.5% of suspects receive guilty verdicts in cases where the witnesses are favorable to the victim.

30The fact that little credence is given to the victim’s word derives from a particular conception of racism. Professional definitions of racist situations are shaped by social reasoning (“common sense”) and emphasize two decisive elements: the suspect’s degree of politicization and the absence of any relationship between the suspect and the victim. The ideal case of racism is political/ideological and unprovoked. One prosecutor gives the example of abuse between two motorists: “They start insulting each other over a parking space, and like idiots they begin to trade insults about the most visible aspects of each person. If one of them [was] one-eyed, he’d get called one-eyed” (interview with a prosecutor). For his deputy, “you have people saying horrible things without necessarily having a racist motive. They say it just as they would anything else” (interview with a deputy prosecutor in charge of racist offenses). Prosecutors are generally convinced that “the genuine racist act is one that has no basis in normal life. . . it’s racism that turns that feeling or opinion. . . into an act” (interview with a prosecutor). The “real” racist is already ideologically or politically racist, and verbally or physically attacks an individual unknown to him or her because of that individual’s real or perceived membership of a racial group. Given this narrow definition of racism, it is unsurprising that the prosecutor is optimistic: “Frankly, genuine racist acts around here over the last five or six years. . . [doubtful intake of breath] I don’t really remember anything that was purely racially motivated” (interview with prosecutor). This reasoning is similar to the general perception in some US police departments that very few crimes are “truly” motivated by racial hatred.  [60] The implicit definition of a “normal” hate crime involves an explicitly racist hate crime, like cross-burnings or assaults by the Ku Klux Klan.

Table 4. Cases by victim-accused relationship

Relationship between victim and accusedNot availableOther grounds for non-prosecutionNon-prosecution: Unverified racist offenseCase prosecutedIncluding case going to courtIncluding guilty verdictTotal suspects
Distant acquaintancen.364137326
No relationshipn.22546523820125
Not availablen.0891910236

Table 4. Cases by victim-accused relationship

a Including relationships with friends, romantic partners, and relatives.
b Including relationships with police or prison officers and civil servants.
Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017. Calculations based on the total number of cases with at least one item of data available on the suspect.

31However, our data show that this ideal type—an extreme right-wing suspect who is completely unknown to the victim—is extremely rare. Of the 458 suspects in our data, only 11 were members of political movements, 9 of them far-right movements. The reality of everyday racism and victims’ experience of it are out of step with this ideal type; the majority of cases involve people who know each other personally, from the neighborhood (29.7% of suspects), from work or school (18.3%), and so on (see Table 4), but such cases are less frequently prosecuted.  [61] We observe a similar process when victim and suspect have a work or school relationship (e.g., between students).  [62]

32Conversely, the absence of such a relationship or the existence of a distant relationship seems to be a factor in favor of prosecution.  [63] The figures are even clearer when victim and accused are “distant acquaintances.”  [64] The only exception is for cases involving administrative relationships (including police and gendarmerie officers, hospital staff, etc.). But this does not mean that cases are always classified in the same way when the victim and accused know each other.  [65] Most of the time, these results involve situations where the victim was an administrative officer—an unfavorable state of affairs for the accused.

33On the other hand, Susan E. Martin’s survey of police handling of hate crimes shows a lower prosecution rate when there is a prior conflict between accused and victim, when the victim “provoked” the accused, or when the victim is “suspect.” The same holds true in France.  [66] This is confirmed by regression analysis: such cases are 2.78 times more likely to be not prosecuted than cases with no prior conflict (see Table 1). Public prosecutors already filter out a vast number of cases, but the court has an even greater effect. This is less pronounced in cases where the victim “provoked” the suspect.  [67] But the effect is far clearer in cases of cross-complaints, where the defendant files a complaint against the victim.  [68]

34These data reveal a crucial feature of prosecutors’ administrative and social reasoning: the “good victim” must be “pure”—that is, he or she must be blameless, with no previous conflict with the accused, no “provocative” behavior, and no tendency to arouse any suspicion in his or her own right.  [69] Otherwise, prosecutors may not prosecute the case on the grounds of “victim behavior”: “‘Victim behavior’ is when there’s something provocative. You hit me, I call you a ‘dirty Arab’ [...] You can be prosecuted for hitting me, and I can be prosecuted for insulting you. [...] It’s more like provocation” (interview with a prosecutor). Such reasons for dismissing the case have been discouraged by the ministry of justice in favor of other grounds, including that the offense is “insufficiently characterized.” This idea of the “good victim” is well illustrated in the remarks of a deputy prosecutor:

Table 5. Cases with confrontational relationships

Confrontational relationshipsNot availableOther grounds for non-prosecutionNon-prosecution: Unverified racist offenseCase prosecutedIncluding case going to courtIncluding guilty verdictTotal suspects
Previous conflictn.95187441913191
Victim’s reactionn.7385738159140

Table 5. Cases with confrontational relationships

Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017. Calculations based on the total number of cases with at least one item of data available on the suspect.


Even though provocation is no excuse, there are people who are pushed to such an extreme that, under certain circumstances, they may well lose their temper a little, or say things they weren’t thinking. [...] We will always try to find another criterion for classifying the case, as it’s difficult [to justify] the victim’s behavior. [...] In terms of violence in general, I won’t deny that when someone really bothers the suspect until he loses his temper, we might say: “Well, he doesn’t deserve to be treated like a good victim.” [...] Next time he’ll think before he gets into trouble. (Interview with a deputy prosecutor in charge of racist offenses)

Recognizing anti-Semitism and Islamophobia

36The type of racism is also a factor in whether cases are unverified. Cases of anti-Semitism are thirty-nine times more likely to be verified than others (see Table 1). Cases of Islamophobia are eight times more likely to be verified than others. These results reveal an implicit hierarchy of racism and antiracism, with anti-Semitism and Islamophobia at the summit, and other forms of racism—particularly anti-North African and anti-black racism—taking a lower position. This is surely the product of a particular sensitivity on the part of prosecutors to anti-Semitism, fostered by frequent directives from the ministry of justice. In turn, these reflect the success of antiracist and/or Jewish organizations in making anti-Semitism a government priority.  [70] It is more difficult to explain prosecutors’ sensitivity to anti-Muslim racism, since anti-Islamophobia organizations are not viewed as legitimate actors by the public authorities.  [71] The fact that treatment differs according to type of racism is generally verified by the way their proportion changes over the course of the legal process (see Table 6), even if there is sometimes a clear difference of opinion between the prosecutor’s office and the court.

Table 6. Cases by type of racism and anti-French hostility

Types of racism and anti-French hostilityNot availableOther grounds for non-prosecutionNon-prosecution: Unverified racist offenseCase prosecutedIncluding case going to courtIncluding guilty verdictTotal suspects
Anti-North Africann.64582632719196
Anti-French hostilityn.1126189137
Implicit racismn.0420006

Table 6. Cases by type of racism and anti-French hostility

Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017. Calculations based on the total number of cases with at least one item of data available on the suspect.

37Indeed, criminal judges seem less inclined to recognize anti-French hostility than prosecutors.  [72] However, the difference between the prosecutor and the court does not hold for the thirty-seven defendants in cases of anti-Semitism.  [73] At the other end of the hierarchy are cases of anti-black racism.  [74] Other forms of racism lie between these two extremes, and the corresponding figures are close to the average for the entire sample—although we must emphasize that sentences for Islamophobia are a special case. Seven of thirty-two defendants (21.9%) were found guilty, revealing that judges are more alert to this type of racism than prosecutors. The data show that the hierarchy of racisms changes during the legal process, and especially between the prosecution and sentencing phases.

“Good victims” and “good suspects”

38The final element to consider is the profile of those involved in racism cases: their gender, age, ethnicity, and criminal record, and whether or not the accused confesses. Cross-referencing these variables allows us to get a clearer picture of the “good victim” (an upper-class white male) and the “good suspect” (a young white or minority working-class male with a criminal record who confesses).

39For victims, while age does not seem to be a determining factor, sex certainly is, although males are over-represented (57.3% of victims, 63.2% of accused). A racist offense is more likely to be verified for male victims than female ones.  [75] The victim’s social class plays a determining role. Although there are gaps in the information available on the class of those involved—recorded for 67.3% of victims and 64.5% of accused—racist offenses are more likely to be verified when the victims are from higher social classes.  [76] The victim’s ethnicity influences the judicial process: paradoxically, black and Arab victims have their cases verified less frequently than white French victims.  [77] Discretion over whether to verify a case as racist typically reproduces inequalities unfavorable to female, non-white, and working-class victims. Sexism, classism, and racism toward victims seemingly influence prosecutors’ decisions, highlighting other aspects of the “good victim”—namely, being an upper-class white male.

Table 7. Profile of the 333 victims in cases where at least one suspect (physical person) was identified

VictimNot availableOther grounds for non-prosecutionNon-prosecution: Unverified racist offenseCase prosecutedIncluding case going to courtIncluding guilty verdictTotal suspects
Not Availablen.0210003
Social classbWorkingn.22848422415120
Not Availablen.7306032126129
Etnic groupcArabn.43570461814155
Not Availablen.0111003

Table 7. Profile of the 333 victims in cases where at least one suspect (physical person) was identified

a At the time of the events.
b Based on declared profession.
c Based on citizenship, place of birth, and the first and last names of individuals and their parents.
Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017. Calculations based on the total number of cases with at least one item of data available on the suspect.

40For suspects, gender is a determining factor. Cases are more likely to be verified for male suspects than female ones.  [78] There is a clear difference between whether a case is prosecuted and whether it goes to court: women appear before court less frequently and are less frequently found guilty. These data are unsurprising, since research on legal action against female offenders has already shown that there are gender-based differences in how they are treated by the legal system, involving an implicitly gendered conception of women: they are responsible for biological reproduction, and so are less likely to be prosecuted or convicted.  [79]

41The suspect’s age seems to be a determining variable, unlike that of the victim. Cases are more likely to be verified for younger suspects than older ones.  [80] The figures are particularly striking for those over the age of sixty. By all appearances, prosecutors and courts are more lenient toward older people. Once again, the accused’s social class is decisive. Upper-class suspects are prosecuted less often than those from the working and middle classes.  [81] The higher he or she is in the social hierarchy, the less likely the suspect is to be prosecuted and convicted, confirming a long tradition of French research into criminal justice.  [82]

Table 8. Profiles of the 458 suspects (physical persons)

SuspectNot availableOther grounds for non-prosecutionNon-prosecution: Unverified racist offenseCase prosecutedIncluding case going to courtIncluding guilty verdictTotal suspects
Not Availablen.1101113
Social classbWorkingn.44180713925196
Not Availablen.115259552812177
Ethnic groupcFrenchn.75481814325223
Not Availablen.113554224

Table 8. Profiles of the 458 suspects (physical persons)

a At the time of the events.
b Based on declared profession.
c Based on citizenship, place of birth, and the first and last names of individuals and their parents.
Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017. Calculations based on the total number of cases with at least one item of data available on the suspect.

42Furthermore, a criminal record is one of the most important contributing factors to a case being verified. Cases of racism in which the suspect has more than one conviction are thirteen times more likely to be verified than those where the suspect has no prior convictions (see Table 1). This indicates a general feature of the justice system and that prosecutors are less tolerant of those who have already been convicted in other cases.  [83] In other words, a suspect is more likely to be considered racist if he or she has already been categorized as an offender by the justice system. In addition to “good victims,” there are also “good suspects.”

43Another factor is whether the accused confesses. Cases in which the accused confessed fully or in part were, respectively, 4.8 and 4.2 times more likely to be verified than cases where the accused denied the charges. Confessing seems to favor prosecution and alternatives to prosecution. This can only be explained by the distinctive features of situations that pit one person’s word against another’s. When there are no witnesses—as was true of 54% of the cases in the regression analysis—only a confession by the suspect can ultimately verify the charge.

44A final factor is the accused’s ethnicity, as minorities are over-represented among suspects at every stage in the legal process (see Table 8). Among suspects, 48.7% were white French, 15.3% were European, 24.5% were Arab, and 4.6% were black. Similar proportions are found at the classification, prosecution, and trial stages. In other words, 64% of the suspects are white and 30.8% non-white. The ethnic diversity of suspects involves three related phenomena that challenge the “classic” pattern of white suspects and non-white victims. The police and judicial authorities record instances of racism regardless of the victim’s profile. The legal view of racism as based on criminal classifications is far removed from—perhaps even antithetical to—the sociological view of racism as based on a power relationship.

45First of all, a significant proportion of cases concern instances of racism between minorities, or even within the same minority group. For example, among cases with a victim and a suspect, and in which the victim is Arab (n=155), 60% involve a white French suspect, 16.1% involve a European suspect, 15.5% involve an Arab suspect, and 1.9% involve a black suspect. Among cases with one victim and one suspect in which the victim is black (n=66), 50% involve a white French suspect, 12.1% involve a European suspect, 27.3% involve an Arab suspect, and 3.0% involve a black suspect. Furthermore, 8.1% of suspects—10.6% of all cases—were involved in anti-French hostility. Because existing legislation protects members of nations (as well as races, ethnic groups, and religions), the police and courts consider anti-French acts to be instances of racism. Of course, such legalism is at odds with the sociological definition of racism as involving power relations between dominant majorities and dominated minorities. Among the cases in which there was a victim and a suspect, and in which the victim was a white French person (n=70), 40% involve Arab suspects, 31.4% involve white French suspects, 11.4% involve European suspects, and 11.4% involve black suspects.

46Lastly, the implementation of antiracism law is not exempt from the general rules of the judicial system, which is characterized by institutional racism. The police and courts typically target minority groups. Because of the different treatment these groups receive on the basis of their ethnicity, they constitute an over-represented “criminal clientele” in French police stations, courts, and prisons.  [84] The situation is similar to that of the United States. Indeed,


the suggestion of racial bias in prosecutorial decision making is troubling in light of evidence that members of traditionally oppressed groups, which hate crime laws were ostensibly enacted to protect, appear to be disproportionately arrested under penalty-enhancement statutes. This can occur because the laws are operationalized without regard to societal power dynamics. In other words, bias crimes that target Whites, heterosexuals, or Protestants are regarded as seriously under these statutes as those targeting traditionally victimized social groups.  [85]

48The law is applied outside any social context, and particularly the dynamic of domination between majority and minority groups. This over-representation can be explained by a number of factors, including the fact that African Americans use “racial terms” more than other groups, that minority groups including African Americans are less likely to lodge complaints, and that the US justice system tends in general to over-criminalize the black population.  [86] Despite differences between the French and US criminal justice systems, minorities continue to receive different treatment in the legal system—even in an area of law that is supposed to protect them.


49Our research sheds light on a twilight zone in the scientific literature on antiracism legislation by exploring how it is implemented in practice by public prosecutors. By cross-referencing qualitative and quantitative data, we have highlighted the categories of administrative and social reasoning used by prosecutors, as well as the factors that determine how they apply their discretionary power. Indeed, a combination of political, legal, organizational, contextual, and other factors contributes to the high rate of unverified racist offenses.

50Our research also confirms the results of US-based research into hate crimes, and of French research into the way other types of offenses are treated by the penal system. The factors that shape the legal process produce a hierarchy of offenses, in which racist offenses are viewed as “peripheral”—except for cases of violence and contempt. Everyday judicial practices reveal a definition of “real” racism as ideological and unprovoked, and of “good,” “pure,” and “non-suspect” victims. This contradicts the daily experience of victims of racism. In spite of legal and institutional differences between France and the United States, there are relatively similar categories of administrative and social reasoning which, paradoxically, mean that minorities are over-represented among the accused.

51By validating our hypotheses, we can understand how the characteristics of the cases and protagonists themselves influence the legal process. This is particularly true for types of racism, insofar as the practices of public prosecutors establish a hierarchy of racisms to be combated, with anti-Semitism and Islamophobia at the top, and anti-North African and anti-black racism at the bottom in spite of the high number of such cases. Statistical analysis also reveals that both victims and suspects are treated differently according to their age, sex, social class, ethnicity, criminal record, and whether they have confessed.

52Finally, these results show that the treatment of racist offenses is constrained by the logic of action that influences public prosecutors’ work, as with any other crime, but that it depends on legal policies to combat racism—as shown by the real effect of focusing on anti-Semitism and Islamophobia—and by the negligible impact of racism specialists in public prosecutors’ offices. These results reveal how difficult it is for the courts to take power relations into account when ruling on individual cases. It is as if the initial objective of the legislation, combating racism, had been depoliticized by the administrative reasoning specific to the legal system, and by social reasoning—i.e., a very restrictive definition of racism that closely resembles the commonsense one.

Appendix 1: Methodology

53 With the agreement of the ministry of justice, we obtained access to 500 court files for racism-related cases heard between 2006 and 2015 in three different criminal courts (C1, C2, and C3).  [87] Starting from an initial list of 833 potential cases (388 in C1, 44 in C2, and 401 in C3), obtained using CASSIOPÉE,  [88] we constructed a sample of 795 cases. For C1 and C2, we tried to be as exhaustive as possible, as the total number of potential cases was relatively small. However, given the limited time available for coding the files, the large number of cases in C3 required us to be selective,  [89] retaining 249 cases (214 non-prosecutions and 35 sentences) out of 401 possible cases. Out of a sample of 795 cases, we processed 500: 275 (55%) in C1, 17 (3.4%) in C2, and 208 (41.6%) in C3. The discrepancy between the number of cases sampled and the number of cases coded is explained by physical loss of files, and the low number of files C2 could make available to us (of 44 sampled, only 17 were found). The overall loss rate was 38.6%, but the rate was low for judgments (almost all of which were found) and especially for dismissed cases, regardless of period, type of case, or reason for non-prosecution (see Table 9). While the discrepancy between reference files (potential cases) and coded files was small, it might constitute a significant bias, making it difficult to interpret the results. We therefore made a statistical adjustment, weighting the cases based on the composition of the reference files (see Table 9).

54We consulted the case files in the archives, using a data entry form created specifically for the study (an interactive Shiny application built using R). The data input form was structured into a “Cases” table and an “Individuals” table (see Table 10), making it possible to code both the “simplest” cases—those with a single victim and no suspect (perpetrator unknown)—and more complex cases involving several individuals. The “Cases” table was used to collect information on the legal process, from the initial record of the incident to any court hearings that took place. The “Individuals” table collected information on the social profile  [90] of suspects, victims, and witnesses, the relationships they may have had with each other, their criminal records, etc. In total, there are 367 variables for the “Cases” table and 103 variables for the “Individuals” table.

Table 9. Composition of the sample

Prosecutor’s decisionN, reference population%, reference populationN, coded files%, coded filesPoint difference between coded files and reference populationN, reference population%, reference populationN, coded files%, coded filesPoint difference between coded files and reference populationN, reference population%, reference populationN, coded files%, coded filesPoint difference between coded files and reference population
Case not prosecuted92.331.1-1.2613.600-13.6 010.50.5
Non-prosecution: No crime215.4155.5024.5211.87.2246115.3-0.7
Non-prosecution: Perpetrator unknown8521.96222.50.61022.715.9-16.852132813.50.5
Non-prosecution: Other nonpenal proceedings or charges10.310.40.136.800-6.8 0211
Non-prosecution: Fault of complainant112.872.5-0.3 000092.241.9-0.3
Non-prosecution: Insufficient charges010.40.4 0000 0000
Non-prosecution: Victim’s behavior102.672.50 000020.5210.5
Non-prosecution: Victim withdraws61.562.20.6 000016452.4-1.6
Non-prosecution: Unfit mental state20.520.70.2 000030.731.40.7
Non-prosecution: Public proceedings halted225.7176.20.5 0000297.2167.70.5
Non-prosecution: Insufficient grounds7619.666244.4715.9317.61.713533.77335.11.4
Non-prosecution: Procedural irregularities0000 00004110.5-0.5
Non-prosecution: Accused cannot be held accountable30.831.10.3 000051.231.40.2
Non-prosecution: Negligible harm20.510.4-0.2 0000174.2115.31
Non-prosecution: Inquiries fruitless20.510.4-0.2 000012394.31.3
Total cases dismissed25064.419269.85.42863.6635.3-28.330876.816981.34.4
Alternatives to prosecution70183713.5-4.6613.6317.645714.22110.1-4.1
Case goes to court6817.54616.7-0.81022.7847.124.3369188.7-0.3
tableau im1
tableau im2

Table 9. Composition of the sample

Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017.

Table 10. Themes and subthemes of data input form tables

Cases tableIndividuals table
PoliceRelationships between actors
ProcedureBiographical information
Classification of incidentType of person
Place and time of incidentMarital status
Descriptions of incidentFamily
Individuals involvedNational/racial group
Public prosecutorHousing situation
Exchanges between police and public prosecutorEducation and profession
Classification of incidentOther information
Public prosecutor’s decisionPolitical and religious life
CourtHarm suffered
HearingLinks with justice system
Public prosecutor’s petitionCriminal record
Legal classificationCustody/hearing
Actors in the hearingDescription of events
Judge’s decisionRecoding of racial categorization
AppealRecoding of situation type

Table 10. Themes and subthemes of data input form tables

Source: “Des paroles et des actes: La justice face aux infractions racistes,” 2017.

55The present article is based on quantitative data from this database. We performed cross tabulations showing significant changes in the type of cases across the four stages of the legal process: the initial classification of the offense, verification of the racist offense by the prosecutor’s office, proceeding to court, and the suspect’s conviction. To compare the beginning and end of the process, we take into account the 458 suspects identified in police reports—that is, those we could find—rather than cases, which gives us a more detailed picture of convictions and acquittals.

56We constructed a specific indicator, the (un)verified racist offense (or [un]verified cases), which is central to our argument. The prosecutor may not prosecute a case for a wide variety of reasons (see the list in Appendix 2), which may be either legal (statute of limitations, withdrawal by the complainant, etc.) or qualitative (assessment of the situation). Our indicator covers non-prosecutions on four grounds: “insufficiently characterized offense,” “no offense,” “victim’s behavior,” and “negligible harm.” In cases involving these grounds, the racist offense was initially classified by the police or gendarmerie, but the prosecutor decided not to pursue it. Conversely, verified racist offenses are cases that are prosecuted or subject to alternatives to prosecution (cautions, mediation, etc.). Unfounded cases cover non-prosecutions for any other reason.

57In this sense, our methodology resembles that of Susan E. Martin.  [91] Unlike Martin, however, we also used logistic regression (see Table 1), which enabled us to cross tabulate the determining factors in discretionary power according to “ceteris paribus” reasoning. We created a unique logistic regression model where the dependent variable is the decision to prosecute a racist offense and where there are multiple independent variables: location, type of offense, type of victim/suspect relationship, type of racism, suspect profile, victim profile, and so on.?  [92] Once again, it was impossible to take all the coded cases into account, since we wanted to test the profile of individual suspects and victims. It was necessary to restrict the sample to the 252 cases involving both a physical victim who filed a complaint and an identified physical suspect, excluding unfounded cases.

58Finally, we conducted twelve interviews with prosecutors and deputy prosecutors. The method used was a semi-structured interview based on a single interview grid, structured around eight themes: the interviewee’s professional background, the organization of the prosecutors’ work, relations between prosecutors and the police or gendarmerie, the legal classification of racist acts, victim profiles, suspect profiles, legal response, and legal proceedings.

Appendix 2: Not prosecuted cases by grounds

Type of caseGrounds for non-prosecutionNumberRemarks and possible subdivisions
Non-prosecutable casesNo crime11Cases which are recorded as criminal, but which prove to be exclusively civil or commercial. Example: a check that is rejected due to insufficient funds.
Insufficient grounds21Vague circumstances, insufficient charges, or insufficient proof.
Legal reasons31 to 37Legal reasons that stop proceedings being undertaken.
31. Public action halted: Complaint withdrawn.
32. Public action halted: Pardon granted
33. Public action halted: Settlement reached
34. Other reasons for halting public action.
35. Immunity
36. Procedural irregularity
37. Accused cannot be held accountable
Prosecutable casesInvestigative problems41 to 4841. Inquiries fruitless
42. Complainant withdraws
43. Deficient mental state
44. Neglect by the complainant
45. Victim’s behavior
46. Victim satisfied, no further reason to prosecute
47. Pretrial remedy
48. Negligible harm caused by offense
Alternative procedures opened by the public prosecutor51 to 5851. Compensation/minor
52. Mediation
53. Therapeutic intervention
54. Complainant satisfied, no further reason to prosecute
55. Remedy on petition of public prosecutor
56. Caution
57. Case referred to health, social, or professional services on petition of public prosecutor
58. Pretrial settlement
Other nonpenal proceedings or charges61Involving cases where a nonpenal response was pursued. An example from the guide published by the chancellery: deporting foreign nationals whose papers are not in order, commercial sanctions for personal bankruptcy, or restrictions on business ownership by commercial courts, settlements by insurance companies for road accidents, etc. Statistically, reason 61 is included in alternatives to prosecution.
Non-prosecutable casesPerpetrator unknown71It was not possible to identify the perpetrator(s) of the offense. Cases with this code are treated automatically as crimes, and are ascribed (sometimes rather too quickly) to “unknown perpetrators.”
Non-prosecution-referred for family monitoring81This reason is included within the No crime category.


  • [1]
    This article is the result of research cofinanced by the Institute for Advanced Studies in the Humanities (University of Edinburgh) and Marie Curie Actions (FP7) as part of the EURIAS Fellowship program. Our thanks to Fabien Jobard, Sébastien Chauvin, Sébastien Delarre, Océane Pérona, Barbara Perry, Neil Chakraborti, Lionel Zevounou, and the three anonymous reviewers for their valuable comments on earlier versions of this text. Any remaining mistakes are our own. We also thank Yassine and Nadia Aslafy, Jérôme Oriol, and Élodie Leconte for their help.
  • [2]
    Erik Bleich and Charles Girard, “Que faire des discours de haine en démocratie?” Esprit 10 (2015): 5-10.
  • [3]
    Ulysse Korolitski, Punir le racisme? Liberté d’expression, démocratie et discours racistes (Paris: CNRS Éditions, 2015).
  • [4]
    Gwénaële Calvès, Envoyer les racistes en prison? Le procès des insulteurs de Christiane Taubira (Issy-les-Moulineaux: LGDJ, 2015).
  • [5]
    Erik Bleich, The Freedom to Be Racist? How the United States and Europe Struggle to Preserve Freedom and Combat Racism (New York: Oxford University Press, 2011).
  • [6]
    Erik Bleich, “Deux poids, deux mesures? La justice française face aux discours islamophobes,” Esprit 10 (2015): 33-44.
  • [7]
    According to Bleich, the “enforcement rate” of racist speech is the rate at which the Court of Cassation verifies the judgment of the Courts of Appeal.
  • [8]
    Bleich, “Deux poids, deux mesures,” 34.
  • [9]
    Michel Wieviorka et al., The Lure of Anti-Semitism: Hatred of Jews in Present-Day France (Leiden: Brill, 2007).
  • [10]
    For instance, Nicolas Herpin, L’application de la loi: Deux poids, deux mesures (Paris: Seuil, 1977); Bruno Aubusson de Carvalay, “Hommes, peines et infractions: la légalité de l’inégalité,” L’Année sociologique 35 (1985): 275-309; René Lévy, Du suspect au coupable: Le travail de police judiciaire (Geneva: Médecine et Hygiène, 1987); Jean Danet, ed., La réponse pénale: Dix ans de traitement des délits (Rennes: Presses Universitaires de Rennes, 2013); Fabien Jobard and Sophie Névanen, “Colour-Tainted Sentencing? Racial Discrimination in Court Sentences Concerning Offenses Committed against Police Officers (1965-2005),” Revue française de sociologie 48, no. 2 (2007): 243-72.
  • [11]
    Maryse Jaspard, Les violences contre les femmes (Paris: La Découverte, 2011).
  • [12]
    Océane Pérona, “La difficile mise en œuvre d’une politique du genre par l’institution policière: Le cas des viols conjugaux,” Champ pénal-Penal Field 14 (2017); Solenne Jouanneau and Anna Matteoli, “Les violences au sein du couple au prisme de la justice familiale: invention et mise en œuvre de l’ordonnance de protection,” Droit et société 99 (2018): 305-21.
  • [13]
    Nathan Hall, “Policing Hate Crime in London and New York City: Some Reflections on the Factors Influencing Effective Law Enforcement, Service Provision and Public Trust and Confidence,” International Review of Victimology 18, no. 1 (2012): 73-87.
  • [14]
    James J. Nolan and Yoshio Akiyama, “An Analysis of Factors That Affect Law Enforcement Participation in Hate Crime Reporting,” Journal of Contemporary Criminal Justice 15, no. 1 (1999): 111-27.
  • [15]
    Samuel Walker and Charles M. Katz, “Less than Meets the Eye: Police Department Bias Crime Units,” American Journal of Police 14, no. 1 (1995): 29-48; Elizabeth A. Boyd, Richard A. Berk, and Karl M. Hamner, “’Motivated by Hatred or Prejudice’: Categorization of Hate-Motivated Crimes in Two Police Divisions,” Law & Society Review 30, no. 4 (1996): 819-50; Shea W. Cronin et al., “Bias-Crime Reporting: Organizational Responses to Ambiguity, Uncertainty, and Infrequency in Eight Police Departments,” American Behavioral Scientist 51, no. 2 (2007): 213-31.
  • [16]
    Susan E. Martin, “’A Cross-Burning Is Not Just an Arson’: Police Social Construction of Hate Crimes in Baltimore County,” Criminology 33, no. 3 (1995): 303-26, 303.
  • [17]
    Eugene H. Czajkoski, “Criminalizing Hate: An Empirical Assessment,” Federal Probation 56, no. 3 (1992): 36-40; Karen Umemoto and C. Kimi Mikami, “A Profile of Race-Bias Hate Crime in Los Angeles County,” Western Criminology Review 2, no. 2 (2000): 1-34; Karen Franklin, “Good Intentions: The Enforcement of Hate Crime Penalty-Enhancement Statutes,” American Behavioral Scientist 46, no. 1 (2002): 154-72.
  • [18]
    Ryken Grattet and Valerie Jenness, “The Reconstitution of Law in Local Settings: Agency Discretion, Ambiguity, and a Surplus of Law in the Policing of Hate Crime,” Law & Society Review 39, no. 4 (2005): 893-942.
  • [19]
    See Bryan D. Byers, Kiesha Warren-Gordon, and James A. Jones, “Predictors of Hate Crime Prosecutions: An Analysis of Data from the National Prosecutors Survey and State-Level Bias Crime Laws,” Race and Justice 2, no. 3 (2012): 203-19, 204.
  • [20]
    Evan M. Read, “Put to the Proof: Evidentiary Considerations in Wisconsin Hate Crime Prosecutions,” Marquette Law Review 89, no. 2 (2005): 453-74.
  • [21]
    Byers, Warren-Gordon, and Jones, “Predictors of Hate Crime Prosecutions.”
  • [22]
    Peter Finn, “Bias Crime: Difficult to Define, Difficult to Prosecute,” Criminal Justice 3 (1988): 19.
  • [23]
    Jeannine Bell, Policing Hatred: Law Enforcement, Civil Rights, and Hate Crime (New York: New York University Press, 2002). Anticipation of judgments by police and prosecutors has also been documented in France (see, for instance, Pérona, “La difficile mise en œuvre”).
  • [24]
    Beverly McPhail and Valerie Jenness, “To Charge or Not to Charge? That Is the Question: The Pursuit of Strategic Advantage in Prosecutional Decision-Making Surrounding Hate Crime,” Journal of Hate Studies 4, no. 1 (2010): 89.
  • [25]
    Nickie D. Phillips, “The Prosecution of Hate Crimes: The Limitations of the Hate Crime Typology,” Journal of Interpersonal Violence 24, no. 5 (2009): 883-905.
  • [26]
    James Garofalo and Susan E. Martin, “Bias Motivated Crimes: Their Characteristic and Law Enforcement Response,” report, Carbondale, Illinois University Center for the Study of Crime (1993).
  • [27]
    Valerie Jenness and Ryken Grattet, “Policing Hate Crime in California,” report, Berkeley, California Policy Research Center (2004).
  • [28]
    Bryan Byers and Richard A. Zeller, “An Examination of Official Hate Crime Offense and Bias Motivation Statistics for 1991-1994,” Journal of Crime and Justice 20, no. 1 (1997): 91-106; Bryan Byers and Richard A. Zeller, “Official Hate Crime Statistics: An Examination of the ‘Epidemic Hypothesis,’” Journal of Crime and Justice 24, no. 2 (2001): 73-85.
  • [29]
    James B. Jacobs and Kimberly Potter, Hate Crimes: Criminal Law & Identity Politics (New York: Oxford University Press, 1998).
  • [30]
    Donald P. Haider-Markel, “Regulating Hate: State and Local Influences on Hate Crime Law Enforcement,” State Politics & Policy Quarterly 2, no. 2 (2002): 126-60.
  • [31]
    Byers, Warren-Gordon, and Jones, “Predictors of Hate Crime Prosecutions.”
  • [32]
    McPhail and Jenness, “To Charge or Not to Charge?”
  • [33]
    Jacobs and Potter, Hate Crimes.
  • [34]
    Haider-Markel, “Regulating Hate.”
  • [35]
    This is the view taken by much French research into other crimes. See in particular Danet, La réponse pénale; Jobard and Névanen, “La couleur du jugement”; and Pérona, “La difficile mise en œuvre.”
  • [36]
    Kimberlé William Crenshaw, “Mapping the Margins: Intersectionality, Identity Politics, and Violence against Women of Color,” Stanford Law Review 43, no. 6 (1991): 1241-99.
  • [37]
    Philomena Essed, Understanding Everyday Racism: An Interdisciplinary Study (Newbury Park: Sage Publications, 1991).
  • [38]
    The legislation in place follows the law of July 1, 1972, which established the crimes of racial abuse, racial defamation, and provocation to racial hatred, and the law of February 3, 2003, which established a racial aggravating circumstance for some crimes (homicide, assault, etc.). We did not take into account the law of January 27, 2017, which extends this to all criminal offenses, as our window of inquiry ends in 2015. As a result, we are interested in crimes committed “because of ethnicity, race, nationality, or religion.” From a purely legal point of view, an act becomes a crime when three elements are present: the legal element (the existence of a legal text criminalizing the act), the material element (the existence of proof), and an intentional element (a crime undertaken deliberately).
  • [39]
    Michael Lipsky, Street-Level Bureaucracy: Dilemmas of the Individual in Public Services (New York: Russell Sage Foundation, 1980).
  • [40]
    Boyd, Berk, and Hamner, “Motivated by Hatred or Prejudice.”
  • [41]
    Alexis Spire, “Histoire et ethnographie d’un sens pratique: le travail bureaucratique des agents du contrôle de l’immigration,” in Observer le travail: Histoire, ethnographie, approches combinées, ed. Yves Cohen et al. (Paris: La Découverte, 2008), 61-76.
  • [42]
    For the construction of this indicator, see Appendix 1, and Narguesse Keyhani, Abdellali Hajjat, and Cécile Rodrigues, “Saisir le racisme par sa pénalisation? Apports et limites d’une analyse fondée sur les dossiers judiciaires,” Genèses: Sciences sociales et histoire 116 (2019): 125-144.
  • [43]
    See Jacobs and Potter, Hate Crimes, 27.
  • [44]
    Hall, “Policing Hate Crime in London and New York City.”
  • [45]
    This variable is explained in Appendix 1.
  • [46]
    Ceteris paribus, with an odds ratio of 0.24, cases in C3 are 4.17 times more likely to be treated as unverified racist offenses than cases in C1.
  • [47]
    This dimension is also present in rape cases. See Sylvie Cromer et al., “Les viols dans la chaîne pénale,” research report, University of Lille Droit et Santé—CRDP/University of Nantes—Droit et Changement Social, 2017.
  • [48]
    Contempt (0%), provocation (18.2%), vandalism (22.2%), violence (25%), defamation (40%), abuse (40.9%), and discrimination (43.5%).
  • [49]
    This phrase is not legally precise, as some of these individuals were subject to “alternatives to prosecution.” We have retained it here for ease of reading.
  • [50]
    Among the 112 suspects charged with racist violence, 25% of cases were unverified, 37.5% were prosecuted, 27.7% appeared before a court, and 13.4% resulted in a guilty verdict. Regression analysis shows that, with an odds ratio of 4.43, cases of racist violence were 4.4 times more likely to be prosecuted than nonviolent cases (see Table 1). By contrast, among the 399 charged with racist speech, 40.9% of cases were unverified, 30.1% were prosecuted, 15.3% went to court, and 9.8% resulted in a guilty verdict.
  • [51]
    Once again, the numbers are low. Among 23 cases of discrimination, 10 were unverified (43.5%), 5 were prosecuted (21.7%), 3 went to court (13%), and 1 resulted in a guilty verdict (4.3%).
  • [52]
    The orders of June 27, 2012, and August 4, 2014, and the ministry of justice’s circular of January 12, 2015, provide for the appointment of a deputy prosecutor in each public prosecutor’s office in charge of cases involving racism.
  • [53]
    Boyd, Berk, and Hamner, “Motivated by Hatred or Prejudice”; Cronin et al., “Bias-Crime Reporting.”
  • [54]
    We find something similar with the idea of “consent” and sexual violence. See Pérona, “La difficile mise en œuvre.”
  • [55]
    Roger Grimshaw and Tony Jefferson, Interpreting Policework: Policy and Practice in Forms of Beat Policing (London: Allen & Unwin, 1987).
  • [56]
    Benjamin Bowling, Violent Racism: Victimization, Policing, and Social Context (Oxford: Clarendon Press, 1998).
  • [57]
    Boyd, Berk, and Hamner, “Motivated by Hatred or Prejudice.”
  • [58]
    Among the 246 suspects in cases without witnesses, 38.6% had unverified cases, 25.6% were prosecuted, 13% went to court, and 8.9% were found guilty (see Table 3). By contrast, among the 88 suspects in cases with one or more witnesses for the victim, 26.1% had unverified cases, 58% were prosecuted, 35.2% went to court, and 20.5% were found guilty.
  • [59]
    Among the 52 suspects in cases with one or more witnesses favorable to the suspect, 46.2% had unverified cases, 21.2% were prosecuted, 11.5% went to court, and only 1.9% resulted in a guilty verdict.
  • [60]
    Boyd, Berk, and Hamner, “Motivated by Hatred or Prejudice.”
  • [61]
    Among the 136 suspects in cases involving neighbors, 42.6% had unverified cases, 25% were prosecuted, 7.4% went to court, and 4.4% were found guilty.
  • [62]
    Among the 84 suspects, 44% had unverified cases, 21.4% were prosecuted, 10.7% went to court, and 8.3% were found guilty.
  • [63]
    Among the 125 suspects in cases where there was no relation between suspect and victim, 36.8% had unverified cases, 41.6% were prosecuted, 30.4% went to court, and 16% were found guilty.
  • [64]
    Among the 26 suspects in cases where they were only distantly acquainted with the victim, only 4 had unverified cases (15.4%), 11 were prosecuted (50%), 26.9% went to court, and 11.5% were found guilty.
  • [65]
    When an administrative relationship existed between victim and suspect, 28% of suspects had unverified cases, 44% were prosecuted, 10 went to court (40%), and 5 were found guilty (20%).
  • [66]
    Among the 191 suspects in cases involving an existing conflict, 45.5% had unverified cases, 23% were prosecuted, 9% went to court, and 6.8% were found guilty (see Table 5).
  • [67]
    Among the 140 suspects in such cases, 40.7% had unverified cases, 27.1% were prosecuted, 10.7% went to court, and 6.4% were found guilty.
  • [68]
    Among the 78 suspects in such cases, 40 had unverified cases (51.3%), 15 were prosecuted (19.2%), 5 went to court (6.4%), and 4 were found guilty (5.1%).
  • [69]
    This categorization of “good victims” resembles those in other cases, including sexual violence (Pérona, “La difficile mise en œuvre”), human trafficking cases (Milena Jakšić, “Devenir victime de la traite: l’épreuve des regards institutionnels,” Actes de la recherche en sciences sociales 198 [2013]: 37-48), asylum seekers (Karen Akoka, “La fabrique du réfugié à l’OFPRA: Du consulat des réfugiés à l’administration des demandeurs d’asile (1952-1992),” PhD diss. in sociology, University of Poitiers, 2012), or emergency aid (Didier Fassin, “La supplique: Stratégies rhétoriques et constructions identitaires dans les demandes d’aide d’urgence,” Annales: Histoire, Sciences sociales 55, no. 5 [2000]: 955-81).
  • [70]
    Samuel Ghiles-Meilhac, “Mesurer l’antisémitisme contemporain: enjeux politiques et méthode scientifique,” Revue d’histoire moderne et contemporaine 62, nos. 2-3 (2015): 201-23.
  • [71]
    See Abdellali Hajjat and Marwan Mohammed, “Le déni de l’islamophobie,” in Islamophobie: Comment les élites françaises fabriquent le “problème musulman” (Paris: La Découverte, 2013), 199-232.
  • [72]
    We use the expression “anti-French hostility” for the sake of sociological coherence; in current French society, one cannot speak of “anti-French racism,” since racism is defined sociologically as a process of division, hierarchization, and discrimination against minorities to the benefit of the majority. Whites remain the majority group in France. The expression does not include racism against minorities from elsewhere in Europe (Spain, Italy, etc.). Among the 37 suspects in cases involving anti-French hostility, 6 had unverified cases (16.2%), 18 were prosecuted (48.6%), 9 went to court (24.3%), and only 1 was ultimately found guilty (2.7%).
  • [73]
    Among the 37 suspects, 9 had unverified cases (24.3%), 16 were prosecuted (43.2%), 11 went to court (29.7%), and 7 were found guilty (18.9%).
  • [74]
    Among the 110 suspects, 42.7% had unverified cases, 29.1% were prosecuted, 19.1% went to court, and 10.9% were found guilty.
  • [75]
    Among the 143 suspects in cases where the victim was a woman, 45.5% had unverified cases (compared to 34.8% when the victim was male), 27.3% were prosecuted (compared to 38%), 14.7% went to court (compared to 19.3%), and 9.1% were found guilty (compared to 11.8%) (see Table 7).
  • [76]
    Among the 31 suspects in cases involving upper-class victims, 6 had unverified cases (19.4%) (compared to 40% for working-class victims), 13 were prosecuted (41.9%) (compared to 35%), 8 went to court (25.8%) (compared to 20%), and 6 were found guilty (19.4%) (compared to 12.5%) (see Table 7).
  • [77]
    Among the 70 suspects in cases where the victim was white and French, 23 had unverified cases (32.9%), 28 were prosecuted (40%), 20 went to court (28.6%), and 7 were found guilty (10%) (see Table 7). Among the 155 and 66 suspects in cases where the victims were Arab or black, respectively, 70 and 25 had unverified cases (45.2% and 37.9%), 46 and 20 were prosecuted (29.7% and 30.3%), 18 and 11 went to court (11.6% and 16.7%), and 14 and 8 were found guilty (9% and 12.1%).
  • [78]
    Among the 290 male suspects, 31% had unverified cases (compared to 49.7% for female suspects), 39% were prosecuted (compared to 23%), 24.5% went to court (compared to 8.5%), and 11.7% were found guilty (compared to 5.5%) (see Table 8).
  • [79]
    Coline Cardi and Geneviève Pruvost, eds., Penser la violence des femmes (Paris: La Découverte, 2017); Kathleen Daly, “Rethinking Judicial Paternalism: Gender, Work-Family Relations, and Sentencing,” Gender & Society 3, no. 1 (1989): 9-36; Jill K. Doerner and Stephen Demuth, “Gender and Sentencing in the Federal Courts: Are Women Treated More Leniently?,” Criminal Justice Policy Review 25, no. 2 (2014): 242-69.
  • [80]
    Among the 135 suspects aged 18-35, 28.9% had unverified cases, 49.9% were prosecuted, 37% went to court, and 14.8% were found guilty (see Table 8). By contrast, among the 213 suspects aged 35-60, 48.8% had unverified cases, 26.3% were prosecuted, 10.8% went to court, and 7% were found guilty.
  • [81]
    Among the 37 upper-class suspects, 16 had unverified cases (43.2%), 8 were prosecuted (21.6%), 7 went to court (18.9%), and 2 were found guilty (5.4%) (see Table 8). Among the 48 middle-class suspects, 17 had unverified cases (35.4%), 18% were prosecuted (37.5%), 12 went to court (25%), and 5 were found guilty (10.4%). Among the 196 working-class suspects, 40.8% had unverified cases, 36.2% were prosecuted, 19.9% went to court, and 12.8% were found guilty.
  • [82]
    Herpin, L’application de la loi; Aubusson de Carvalay, “Hommes, peines et infractions”; Lévy, Du suspect au coupable; Danet, La réponse pénale.
  • [83]
    Danet, La réponse pénale.
  • [84]
    Jobard and Névanen, “La couleur du jugement”; Fabien Jobard and René Lévy, Police et minorités visibles: Les contrôles d’identité à Paris (Paris: Open Society Justice Initiative, 2009).
  • [85]
    See Franklin, “Good Intentions,” 159.
  • [86]
    It is no coincidence that African Americans are over-represented as perpetrators of hate crimes in these statistics. In 1993, according to national data from the Southern Pacific Leadership Conference, 46% of racially motivated murders were committed by black people. In 1999, according to FBI data, 19% of known perpetrators of racist hate crimes were African American, and 18% of race-based hate crimes were antiwhite hate crimes. The same asymmetry holds for arrest rates, both at local and national levels. In Florida, the distribution of victims (50% white, 38% black) is unrelated to that of perpetrators (33% white, 27% black, 40% unknown). In 1999, California recorded 1,321 racist incidents, 15% of which were antiwhite.
  • [87]
    We contacted fourteen courts, but only three agreed to give us access to their files. In order to preserve their anonymity, we cannot indicate their location. C3 covers the largest population, followed by C1 and, finally, C2. This explains the higher number of cases in C3.
  • [88]
    The use of this tool, and of the 160 NATINF (nature de l’infraction) crime codes for racism cases, is the only way of identifying the prosecutor’s numbers used to classify archived files. However, this produces two sampling biases: (1) CASSIOPÉE only began working well in 2012; (2) cases only appear in our sample if they were registered with a NATINF code for racist offenses. However, it is known that the police did not use these codes when recording certain racism cases, which consequently escaped our notice.
  • [89]
    The principles of selection were as follows: total number of judgments and alternatives to prosecution; total number of cases dismissed because they involved negligible harms, because of the victim’s behavior, and for reasons of mental health; other dismissed cases chosen at random (one in two).
  • [90]
    For social class, we decoded the professions indicated in police reports, which are self-declared (and therefore impossible to verify), and then matched these professions to INSEE’s socioprofessional categories, sorting these into three large groups: working class, middle class, and upper class. For ethnicity, we distinguished between three forms of racial categorization: assigned categorization (where an individual is assigned to a group), stated categorization (where an individual self-identifies with a group), and “informational” categorization (where an individual is identified with a group based on the combination of first and second name, place of birth, nationality, and parents’ names). The third of these is used in the present article.
  • [91]
    Martin, “A Cross-Burning Is Not Just an Arson”.
  • [92]
    The age of the victim and the accused do not produce significant results and have not been included in the model. The logistic regression is based on a selection of cases in which the victim brought charges and the victim and the suspect are unique physical persons, excluding unfounded cases.

This article examines the judicial treatment of offences of racial aggression and the phenomenon of attrition, through an investigation based on the case files of 500 racial hate crimes processed between 2006 and 2015 in three criminal courts in France. It shows that a combination of factors produce a high rate of case dismissal for unconfirmed racist offences, based on a hierarchy of infractions and types of racism, the differential treatment of victims and defendants (according to age, sex, social class, ethnicity, etc.), and the paradoxical over-representation of minorities among the suspects. Despite the legal and institutional differences between France and the United States, both countries draw on relatively similar forms of administrative and social comprehension.


  • justice
  • prosecution
  • discretion
  • hate crime
  • racism
  • quantitative method
Abdellali Hajjat
Abdellali Hajjat is associate professor of sociology at the Université libre de Bruxelles and member of the Group for research on Ethnic Relations, Migration and Equality. His work focuses on Islamophobia and the construction of the “Muslim problem,” antiracism policies, and experiences of racism in the academic world. He discussed these topics in Islamophobie: Comment les élites françaises fabriquent le problème musulman (Paris: La Découverte, 2013), coauthored with Marwan Mohammed; La Marche pour l’égalité et contre le racisme (Paris: Amsterdam, 2013); and, with Julien Beaugé, “Élites françaises et construction du ‘problème musulman’: le cas du Haut Conseil à l’intégration (1989-2012),” Sociologie 5, no. 1 (2014): 31-59 (Université libre de Bruxelles, GERME, CP 124, Avenue Jeanne 44, 1050 Bruxelles).
Narguesse Keyhani
Narguesse Keyhani is associate professor of political science at the University of Lyon 2 and member of Triangle. Having completed a thesis on intercultural relations in France, she went on to join the “Des paroles et des actes” research project (GIP-Justice) as a postdoctoral student. With Vincent-Arnaud Chappe, she is conducting a study on antidiscrimination activism in the SNCF. Her recent publications include: “Agir sur l’opinion. Socio-histoire d’un répertoire d’action antiraciste,” Terrains & Travaux 32 (2018): 201-23; and, with Vincent-Arnaud Chappe, “La fabrique d’un collectif judiciaire: La mobilisation des cheminots marocains contre les discriminations à la SNCF,” Revue française de science politique 68, no. 1 (February 2018): 7-29. She is a former postdoctoral researcher at the Convergences Migrations-INED Institute (Université Lyon 2, UFR ASSP, 5, avenue Pierre Mendès France, 69676 Bron).
Cécile Rodrigues
Cécile Rodrigues is a CNRS engineer in quantitative data production and analysis at the Center for Administrative, Political, and Social Research (CERAPS, University of Lille, CNRS, UMR 8026, 1 Place Déliot, CS 10629, 59024 Lille Cedex).
Uploaded on on 18/12/2020
Distribution électronique pour Presses de Sciences Po © Presses de Sciences Po. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
Loading... Please wait