1 According to the interactionist paradigm, the facade of unity of professions, provided by legal authorizations to practice, corporatism, professional associations, codes of ethics, “licenses” and “mandates” (Hughes 1996), does not constitute proof of homogeneity of their members, who belong to specialities, to competitive professional “segments,” and maintain distinct relationships to their profession according to the status assigned to them and their career stage (Strauss 1992). This heterogeneity gives rise to a plurality of conceptions of the profession which confront each other in the professional arena or beyond giving rise to debates which we call “ideological” in the manner of Strauss, for whom “professional identity can be considered as a counterpart of the ideology of a political movement” (ibid.: 83).
2 The great strength of the Straussian definition of professional arenas as political spaces for debate is that it does not reduce the field of politics to political parties, to professional politicians, to public opinion or to the media. The professions form a “social world” composed of an arena in which points of view about ways of doing things and the place of the profession within society are debated. In the police, these confrontations between divergent points of view and the political dimension of the debate are immediately perceptible because the nature itself of the missions assigned to the police (prevention vs. repression, community policing vs. policing to maintain order) is directly debated in the National Assembly and the Senate and accross the dividing lines between the political parties, between “right-wing” and “left-wing” conceptions of the police and its missions.
3 These debates within the police take an institutionalized form: the number of professional trade unions and their very strong representativeness (police officers are among the most unionized public officials) bears witness to the strength of the internal debates within the profession. Trade union elections in the police in fact constitute an observatory for the adherence of police officers to governmental politics.
4 Analysis of professional ideologies should not be limited to parliamentary, political or trade union arbitrations or formalized and publicized moments of collective action either. It can also be done at a more microscopic level, at the scene of the professional activity itself—as can be seen in other comparable professions, such as fire fighters (Pudal 2011)—, where differences of opinion take shape between colleagues (police officers) and “outsiders” (the non-police) contributing to the creation of a deliberative sociability. This type of localized exchanges is the breading ground itself of democratic public debate in that it makes “an agnostic pluralism” possible (Mouffe 1993). However, it is still necessary that all political positions can be expressed in the workplace.
5 Police officers who possess a trade union mandate have this opportunity. They are not subjected to a duty of confidentiality and express themselves freely in the media. But this is not the case for other police officials, who must seek ministerial authorization to express themselves publically. Between colleagues, are police officers muzzled all the same? The police organization has extensive “channels,” hidden from view from the public and the hierarchy, which allow for quite a high level of release and partially lifts the duty of confidentiality between police officers of the same rank (Pruvost 2008). Without necessarily indicating an affiliation to a particular political party, police officers express maxims that are expressions of general points of views on the management of criminality, work methods, various Interior Ministers and government policies. They implicitly or explicitly clash with opposing partisan ideologies that mark the “politicization” of the debate (Duchesne and Haegel 2004; Braconnier 2010).
6 Nearly twenty years after the pioneering work of Monjardet and Gorgeon (1993, 1996, 1999, 2004) and based on a questionnaire survey of 5,221 police officers of all ranks (Coulangeon, Pruvost and Roharik 2003), we describe the ideological dissensions on how police work should be carried out in 2003. This date is significant: between 2000 and 2002 Jean-Pierre Chevènement  launched Police de Proximité [community policing]. On the 21st April 2002, Jean-Marie Le Pen  qualified for the second round of the 2002 presidential election, three weeks later Nicolas Sarkozy was named Minister of the Interior in the Raffarin government. He then brought an end to Police de Proximité and was, at the time, very favourably received by police officers who were very critical of the reform. Many events could partially explain the results obtained, which are explored in two stages. We will firstly show that the political positioning of police officers is a blind spot for most large-scale surveys of policing, interest in which has however been shown by ethnographers. Next we will study the distribution of police officers in three opinion classes—repressive, preventive and median—defined by means of a latent class analysis carried out on a series of questions on attitudes and opinions contained in the survey questionnaire.
BOX 1.–Outline of the survey
For each police directorate we determined which were the large, small and medium sized divisions from which a random selection was made in proportion to their relative weight in the whole. The respondents were known to be representative in terms of rank and sex in proportion to the total population.
To clarify the results and put them in perspective, we made use of qualitative data (life stories and observations) gathered from a survey on working conditions and the feminization of the police at the same time as the questionnaire survey (Pruvost 2007b).
7 The distribution of political attitudes of occupational groups has not generally been much explored, either by the sociology of occupations or by electoral sociology. The sociology of occupations has been more concerned with the formal involvement of professionals with the reforms that concern them or in collective action (strikes, trade union actions or the creation of professional associations) taken against these same reforms (Kergoat et al. 1992; Israël 2009) than with the intermediate level of day-to-day forms of expression (workplace conversations, for example) and political involvement in a professional context. In electoral sociology, most surveys rely on the occupational and socio-occupational nomenclature of INSEE [French National Institute of Statistics and Economic Studies] (Cautrès 2001; Mayer 2010) and permit a distinction between the public and private sector to be made (Rouban 1999), but limited samples mean that more often than not it is not possible to make arguments at the more precise level of occupational groups.
8 In Anglo-Saxon and Francophone sociology of the police, with the exception of an Australian researcher (Chan, Devery and Doran 2003), the questions of precise political affiliations has not often been asked because of the closedness of the domain itself. At all levels, police officers have a tendency to hide behind their duty of confidentiality: lower ranks are more anxious to preserve confidentiality on a particular subject when the police hierarchy has direct ties to the local, federal or national political majority. In the same way, neither does the police hierarchy wish to have police officers accused of any connivance, large or small, with a particular party. It follows that voting by police officers has not in itself been the subject of a detailed sociological investigation, either statistical or ethnographic. The absence of sociological studies specifically on the political positions of police officers has barely been compensated for by monographic studies that only approach this question at the margins apart from a study carried out on the rank of commissaire [superintendent] (Ocqueteau 2006).
9 This is why we decided to rely on our own observations (Pruvost 2007a) in order to take into account the effects of the duty of confidentiality and take the opportunity to evade them. The norm (at least from what we could see between 1999 and 2005) remains that of the very classic “spiral of silence” that leads to an alignment of opinions to what each believed to be the dominant police doxa in order to mitigate conflicts (Noelle-Neumann 1984). This spiral of silence can, certainly, be curbed by the “big mouths” in the commissariats who comment on the political situation and loudly and forcefully proclaim their opinions, but police officers largely remain cautious: with considerable staff rotation and strong superior officers, political sympathies are more often expressed indirectly through conceptions of the profession. As a result it is difficult to determine the moments (during a career or political life), the type of relations with colleagues, the types of police division, position or work situation that allow some police officers to display their political colours (and to restrain others from doing so at the same time). Trade union affiliations, as we will see, do not indeed constitute very reliable indicators.
10 A case study, that came from an observation carried out in April 2002 in a Sécurité Publique commissariat in the Paris region, at the same time that Jean-Marie Le Pen found himself in the second round of the presidential election, proves to be very instructive in this respect. This event seems indeed to have loosened tongues and most of the police officers observed dropped their normal political reserve. We were thus able to look at four levels: the conversations about disorder held by some gardiens de la paix [police constables/patrol officers] before 21st April 2002; their professional practices in situations (as observed), their reputations within the commissariat; finally how they voted in the first round of the 2002 presidential election, or at least how they a posteriori stated they had voted when in the commissariat.
11 The first level (statement of professional ideologies) will be explored based on a conversation, prior to 21st April 2002, on the appropriateness of the use of tear gas.
After an operation on the public highway, the PS4 squad [Police Secours] [emergency police]  returned to their post to do their paperwork. It is 8.00 pm. The squad takes the opportunity to eat a little and goes down to the kitchen where they come across Jean-Jacques (gardiende la paix, 50, in theBAC [Brigade Anti-Criminalité] [anti-crime squad]). He had arrived early. He is going on duty at 8.30 pm. He is taking advantage of this to have a coffee and to join in with the “gossip.” He explains that “there is a new deputy to the DDSP [Directeur Départemental de la Sécurité Publique] [departmental director of public security]” and the deputy “does not like lacrymos [tear gas].”
Gérard (sergeant in charge of the PS4, 45) replies: “In any case, tear gas does not go down well with the population we have.”
Sonia (gardienne de la paix in the PS4, 35) does not agree: “Even so it does make it simpler to disperse youth.”
Luc (gardien de la paix in the PS4, 32) causes ripples, saying “we need a new ‘Honneur de la Police’.” 
Gérard replies to her smiling: “Isn’t that a bit far right?”
Nobody replies. Jean-Jacques puts his “lunchbox” in the fridge.
Luc explains to me in private that Jean-Jacques eats better than anyone in the commissariat. “He eats organic” (extract from fieldwork journal, 15th April 2002).
13 In the conversation, lasting a few minutes, political positions are elusive. They were clarified after 21st April 2002. Let us detail the positions supported by the protagonists in this dialogue in light of two other levels of analysis: voting in the presidential election and the reputation of the officers.
14 Jean-Jacques does not take sides in the debate on tear gas, he contents himself with passing on the news of the ban on the usage of tear gas, but his silence is eloquent. It seems clear to everyone that he approves the ban. He is renowned for his hostility to the “cowboy” attitudes of some police officers and regularly cites his origins (Polish, formerly working on the production line in the car industry) to justify his different approach to the job. He says, without irony, to his colleagues who had voted for Le Pen that he “had shed a tear of conscience when he saw that Jospin had not made the second round,” without, however, saying whom he had voted for.  On the level of police work, this “burly 1.9m tall man” has a good reputation in the commissariat. He is useful in operations involving young people because his size impresses and “immediately calms the situation,” according to his colleagues. The BAC to which Jean-Jacques belongs would, however, go on to make abusive use of tear gas a few days after the conversation of the 15th April during an operation in a student cafe. The “gassing” victims wanted to make a complaint. Jean-Jacques received the ulcerated students warmly, while remaining in solidarity with his colleagues: he would not accept their complaint.
15 Gérard, who takes a moderate position on the use of tear gas (he advises against its use on the estates in the area) is considered an irreproachable and honest police officer. On operations he does not hide his dislike of the “crassoux” [youth from the estates] and the social role that they too often play, but he deals very properly with the small jobs given to his squad, which has the reputation for being the most reliable in the commissariat. He voted for Le Pen, not because he was a Le Pen supporter, but to “make a point.” He belongs to the Alliance union (positioned on the right).
16 Sonia champions a harder line and advocates the use of arms (she would support this point of view on arms on another occasion during the observation), but within the commissariat she has a reputation as a consensual woman who “doesn’t cause a fuss.” The head of the division would name her as the person responsible for social affairs in April: if there is a problem with a police officer who is performing badly she is supposed to mediate with the hierarchy. She would be furious with this nomination as she does not want to be considered as the social worker in the division, but she is forced to accept the mission by the hierarchy. Sonia also voted for Le Pen to send the following message to the right: “more repression is needed.” She is a member of the Syndicat Général de la Police trade union, which is affiliated to Force Ouvrière (positioned on the left), because in Paris, where she was before, “it was the most powerful.”
17 As for Luc, who is younger than Sonia, he does not hide his racism, which he portrays as “informed” (he has read books on Africa and is educated on sub-Saharan immigration). He voted for Le Pen and this was not the first time he had voted for the Front National leader. He is very legalistic (applying procedures to the letter). His arm of choice: fine stamps. His squad leader (Gérard) has to rein in his fining frenzies to avoid an administrative work overload. Luc is a comic character within the station. He always has anecdotes to tell—funny stories or police dramas in which he portrays himself as the hero or antihero. He is not a member of the FPIP (Fédération Professionnelle Indépendante de la Police, a trade union close to the Front National) because he wants a transfer and he fears that this could be a negative point “in the case of something happening on the VP [voie publique] [the street].” He is in Alliance.
18 Four conclusions can be drawn from this observation in a Sécurité Publique commissariat in April 2002: political and trade unions positions can be unconnected; political positioning does not have a systematic impact on concrete practices or reputations for being a “good police officer;” police officers who voted for Le Pen did not hide this fact, while officers who voted on the left did not explicitly justify it; finally, being positioned on the left does not have to contravene the corporatist solidarity process when there is police malpractice.
19 Gaining access to all these levels—from conceptions of the profession to practices by way of political positioning—is nevertheless uncommon and is partly due to the shock provoked by the second round of the 2002 presidential election. It remains for sociologists to explore the “intermediate” level of conceptions of police work—intermediate in the sense that it is distinguished from publically stated opinions (political, trade union and in the media) and from positions taken at the polls.
Exploring professional ideologies
20 In Anglo-Saxon sociology of the police, conceptions of the job constitute a completely separate field of research, brought together under the heading of the study of “police culture” (Waddington 1999; Westmarland 2008) or the “relationship to ethics” (Klockars et al. 2000; Neyroud 2008). Sometimes quantitative, sometimes qualitative, these studies attempt to define police profiles: authoritarian, racist, cynical, sexist; and lean more towards a quite monolithic profile with positioning that is conservative, if not reactionary (Skolnick 1966). In France, Monjardet has sought, in contrast, to establish the variability of ideological adherences by means of a longitudinal study of a cohort of gardiens de la paix (the lower grade of police officers) (Monjardet and Gorgeon 1993, 1996, 1999, 2004), which allows him to identify the effects of professional socialization over ten years on conceptions of the profession (Alain 2004; Alain and Grégoire 2007).
21 Beyond the debate between the proponents of a “single” police culture and those who place greater emphasis on its pluralities, these studies are based on two styles of investigation by offering a choice between various conceptions of the police and practical cases to resolve in the same questionnaire, which allows the degree of legalism of officers to be tested and correlations, for example between a repressive conception of the profession and a repressive and illegal use of arms in fictional situations presented in the questionnaire, to be established (Monjardet and Gorgeon 1993; Klockars et al. 2000; Weisburg et al. 2001; Alain 2004; Alain and Grégoire 2007).
22 In the following section we rely on data from a socio-demographic survey on working and employment conditions of the police that we conducted in 2003 (Coulangeon, Pruvost and Roharik 2003), in part replicating a statistical survey from the 1980s (Hauser and Masingue 1983). Although the purpose of the survey was not to identify a police culture or police ethics as such (based in particular on the resolving of practical cases), it did comprise four questions, including in total twenty possible responses, which allow coherent conceptions of the profession to be determined. The four questions were as follows: what is the primary mission of the police? To reduce criminality, which partner should the police work with? In your opinion, what are the causes of delinquency? Against which category of people should most vigilance be shown? This last question is distinguished from the three others in the sense that it allows us to delve into the more classical ideological positions set out in the previous questions in more detail by putting monitoring of non-delinquent categories such as youth and immigrants and categories of offenders such as gangsters and pimps on the same level (Pruvost and Névanen 2009).
Latent class analysis
23 To bring to light the ideological attitudes underlying the positioning on these four indicators, we used a method of classification that allows the definition of subpopulations choosing the same response-types on a number of questions. Of the methods that could be used when a population is described by a series or categorical variables, we used latent class analysis (Box 2), which provides a robust decision criterion to retain the number of classes, in contrast to classification by classification tree, usually used as a tool for clustering, which provides the researcher with freedom but also difficulties in choosing the extent of this typology. In latent class analysis, the number of classes is determined on the basis of a statistical test that measures the degree of fit between reality, such as it can be organized by the theoretical model applied, and the reality of the data studied. A too small a degree of correlation implies the rejection of the classification solution estimated by the modelling algorithms.
24 Added to this is the principle of parsimony, which guides decisions when there appear to be multiple solutions to the classification. It is obvious that the greater the number of classes of typologies that are obtained through modelling the more the degree of adequacy is improved. The principle of parsimony is thus used to retain the most economical solution, the one that presents the fewest number of classes in order to maintain a satisfactory level of statistical adequacy.
25 Thus, latent class analysis is a very effective tool for exploring data (of the same order as multiple correspondence analysis, for example) coupled with a better management of the interpretations of facts through statistical tests that it deploys.
BOX 2.–Latent class analysis
In France, latent class analysis is not as popular as it is in the Anglo-Saxon world, despite its precise and detailed presentation, directly inspired by Lazarsfeld, by Raymond Boudon in a Revue Française de Sociologie article in 1962. Indeed, multiple correspondence analysis and methods of classification in a tree structure were more widely used by French sociologists.
From a practical point of view, latent class analysis permits the determination of a collection of latent classes that are reciprocally exclusive ( “locally independent”) based on a cross-distribution of categorical variables. The n latent classes correspond to the n modalities of the non-observed (or latent) categorical variable. This variable can be described by examining the various patterns of interrelations between the observed indicators. Estimation algorithms are then used, from which two types of parameters are identified, which permit the precise description of the latent variable: the probability of latent classes (meaning the probability that a given individual belongs to one, and only one, of the k-classes of the latent variable; in the final description of the latent variable this translates as an indication of the size of each class within the studied population) and the conditional probabilities (which are the respective probabilities that a given individual, belonging to one of the k-classes identified in the latent variable, is characterized by a given modality of the observed variable used in the modelling).
The estimation of these parameters is made by using the maximum likelihood estimation method. The basic idea of maximum likelihood estimation is to quantify the correlation between a probability distribution and sample: the greater the likelihood of the sample, the better the fit. From a practical point of view, this can be done via the EM algorithm (McLachlan and Krishnan 2008). This iterative method permits convergence on the maximum local likelihood and is carried out in two steps: step E (expectation) and step M (maximization). The first step consists of calculating the expectation of the complete log-likelihood (that is to say, assuming the unobserved known classes) and the second of finding the local maximum. The algorithm stops when the difference between the expectation of two complete consecutive log-likelihoods is less than a set tolerance threshold. We speak of a “local maximum” because the likelihood function can have several maxima. Thus, when both of these have been identified, only the greater of the two should be retained. Note, however, that the probability that the likelihood function has a single maximum tends to 1 when the sample size tends to infinity.
In this way we can identify several latent variables with a distinct number of modalities (or latent classes in the terminology of the analysis), and the problem that arises at this stage of the modelling is the choice, among all the hypothetical (latent) variables of the one that best fits the observed data. This choice is guided by a series of statistics, called goodness of fit, which can obtain this or solutions that are the closest to the observed reality from a statistical point of view: BIC (Schwartz’s Bayesian information criterion), AIC (Akaike information criterion), Chi² or G² (maximum likelihood statistical significance test, often presented as L² or LL). Recourse to one or other of these tests is dictated by the size of the sample studied and by certain conditions of the regularity of the distributions. While their interpretation is different, all these tests are consistent with a measurement of the difference between the distributions provided by the latent variables identified and those directly measured through the observed indicators. In our case, we use the LL test, which is more intuitive and adapted to the size of the study population than the Chi2 test (Heinen 1996), together with the BIC criterion that takes account of the parsimony of the model.
While the goodness of fit tests demonstrate that there are several possible solutions, the choice of the “best” model is then guided by the criteria of parsimony, which leads to favouring a reduced number of classes, ensuring a simplified picture of reality and guaranteeing a distribution that offers a sufficiently large number of members to permit further modelling (a typology is rarely the ultimate goal of research, the explanation of social phenomena often leads to the use of a classification as a part-complete indicator in other types of analysis, where there is a need to not “crumble” the numbers so that the scope of other secondary analyses is not compromised).
26 The individuals were then grouped by our model’s algorithm in terms of the observed characteristics they share. The implementation of this method showed that the classification model that best met the two given methodological criteria—quality of fit and parsimony—consisted of three latent opinion classes (Table 1). The usual criterion of goodness of fit leads us, indeed, to accept either the 3-class model (L² = 598.19, df  = 549, p = 0.072), or one with 4 classes (L² = 547.55, df = 532, p = 0.31). The model with five classes does not provide a unique solution (we can not identify a global maximum among the local maxima indentified by the EM algorithm and we thus consider that the model is not stable and reject it), a risk, moreover, that is incurred when the model is estimated with more and more latent classes (Table 1). Then, between the two identified classifications, the model with 3 latent classes and the one with 4 latent classes, the principle of parsimony led us to use the most economical model, made up of 3 distinct profiles (those for which the BIC has the lowest value).
Choice of model
|L² (or LL)||BIC||df||p-value|
|Model 1||1 class||1,091.2042||- 3,875.9496||583||2.7e-33|
|Model 2||2 classes||677.48370||- 4,144.8302||566||0.00085|
|Model 3||3 classes||598.19653||- 4,079.2776||549||0.072|
|Model 4||4 classes||547.55965||- 3,985.0747||532||0.31|
Choice of model
The emergence of three opinion classes: median, repressive and preventive
27 Latent class analysis brings out the strong coherence in the responses of police officers and distinguishes three significantly different classes of police opinion. In their longitudinal study, Monjardet and Gorgeon (1993, 1996, 1999, 2004) for their part distinguished four types of police officer (the open legalists, the closed legalists, the open illegalists and closed illegalists) by combining the degree of legalism with openness to the outside world, which permitted them to confirm the plurality of relationships with the job in a reputedly homogenous profession. A limitation of this division into four groups is that it is based on samples that become insignificant with the reduction in response rate with each repetition of the questionnaire. Alain (2004), Alain and Grégoire (2007) took care to subdivide his sample into only three groups, comprising a group described as “relativist,”  which resembles the “median” category that we created in our own analysis.
28 What indeed should we name the three groups that emerged from the latent class statistical method? This is a significant taxonomic issue, since sociological analysis is guided by the formulations chosen. We opted for a police terminology, adopting the terms repressive and preventive for the first two groups, and we called the intermediate group median (Table 2).
|Primary mission of the police||Arrest delinquents||24||5.6||4.4||13.9|
|Ensure respect for the law||57.4||58.9||39.6||56.3|
|Rescue people in danger||10.7||13.6||18.5||12.8|
|Protect republican institutions||5||9.8||24.3||9.1|
|To reduce criminality, the police must collaborate more closely with||Parents||1.6||27.0||1||12.9|
|Partners in civil society (doctors, associations, teachers)||5.7||28.4||68.2||22.1|
|The increase in delinquency is primarily due to||Poverty and unemployment||5.6||10.3||71.4||14.3|
|A lack of supervision in the family or at school||26.7||86.9||1.8||51.1|
|A lack of sanctions in the justice system and prison places||61.6||2.1||14.9||30.3|
|The black economy||4.8||0||11.3||3.3|
|The fact there is too much crime in films and the media||1.4||0.6||0.6||1|
|With respect to which categories do you think most vigilance should be shown?||Youth||10.5||31.2||4.6||19.2|
|Gangsters, pimps and prostitution**||19.0||5.4||19.3||12.9|
|Drugs (drug traffickers, users)||29.5||30.2||27.0||29.6|
** We combined the pimps and prostitutes category with that of gangsters because the responses were too small to be separate response categories.
29 Before showing the clear distinctions between the three classes of police officer, it is worth noting that there are only two responses out of the twenty possible with similar results for the repressive, median and preventive classes: these are, on the one hand, vigilance with respect to drug dealers and users that united a little less than a third of each opinion class, and, on the other, films and the media, that only around 1 % of police officers in each of the three classes identified as a cause of the increase in delinquency.
30 There are no other responses that received the same response rates from the three opinion classes—a sign that divergent points of view do indeed exist.
31 The medians make up 45 % of our sample.  We have called them the medians because they embody the middle road in the statistical sense of the term and the police doxa in the sociological sense. The medians are also characterized by their intermediate place between the two extremes, which are the supporters of a repressive police or a preventive police, either in the sense that they show similar results to one or other of the groups or that they follow the same tendency, but to a lesser extent.
32 Their uniqueness in comparison to the two other groups concerns relationships to parents and to youth: 27 % of them believe that there should be closer collaboration with parents, while this dimension is totally negligible for the repressive and preventive groups (1.6 % and 1 % respectively). In the same vein, the medians believe that vigilance should be shown towards youth in particular (up to 31.2 % of them)—a vigilance that very few of the preventive (4.6 %) and a medium number of the repressive group (10.5 %) show interest in.
33 This attention shown to youth is put into context by the variable that distinguishes the medians most from the other opinion classes: 86.9 % of them consider that delinquency is due to a lack of supervision in the family and at school (in contrast to the preventives, of whom only up to 1.8 % cite this cause).
34 In all cases, the primary choice of response by the medians is coherent and corresponds to the pragmatism that is a characteristic of ordinary police officers: they take into account the most immediately accessible populations and with whom they are confronted most frequently in the commissariats, namely parents and youth.
35 The medians come closer to adherents of preventive policing in the distribution of data on the priority missions of the police: as opposed to the repressives, a small number of them selected the rescue (13.6 %) and prevention of delinquency (12.1 %) options. The second rate is similar to within about a point to the preventives, the other is close to a lesser degree.
36 The medians come closer to the adherents of repressive policing on one point: 58.9 % of them believe that the primary mission of the police is to ensure respect for the law, but this is the only close result that can be detected. The medians are more often to be found following a middle way between the repressive and preventive groups, sometimes closer to one or the other: thus protection of republican institutions seems to be the priority mission for 9.8 % of them (i.e., 5 points more than the repressives and 14 less than preventives). For 35.7 %, collaboration with gendarmes seems to be important (10 points less than the repressives and 7 points more than the preventives). Collaboration with civil society also seems to them to be justified for 28.4 % (23 points more than the repressives and 40 points less than the preventives). In the same way, 10.3 % of them believed that delinquency was due to socio-economic conditions (5 points more than the repressives, and 61 less than the preventives). As for vigilance towards car drivers, the medians occupy an intermediate position between the repressives and preventives.
37 The medians thus share a mixture of repressive and preventive points of view, exhibiting a distribution with no extreme, with the exception of everything that concerns questions concerning families—embodying a sort of doxic relativism that reflects the point of view of ordinary police officers well, navigating between contradictory injunctions. This group cites the responsibilities of parents more than social inequalities.
Supporters of repressive policing
38 This group accounts for 45 % of the total, the same level as the median group. It is characterized by a propensity to consider that the missions of the police are to ensure respect for the law (57.4 %) and exhibits a different distribution of other priority missions to the other two groups. Indeed, in second place is the mission to arrest delinquents (24 % against 5.6 % for the medians and 4.4 % for those who adhere to preventive policing). The protection of republican institutions and the prevention of delinquency are, moreover, only considered primary by a small minority of them.
39 To reduce criminality, the repressives choose overwhelmingly to collaborate equally with magistrates (46.7 % of them against 8.8 % of medians and 1.8 % of the preventives) and gendarmes (45.9 % of them against 35.7 % of medians and 29 % of the preventives). Conversely, only 5.7 % and 1.6 % of adherents of repressive policing believe that closer work with partners in civil society and parents is needed.
40 In the same vein, the majority of supporters of repressive policing believe that the increase in criminality is due to the lack of judicial sanctions and the lack of prison places (61.6 % of them). M. Gardib, gardien de la paix, is completely representative of this position:
My complaint is that the problems stem from the justice system. It needs to be better applied. If the pros like us have problems too, it’s because it’s a complete mindset. You can do what you want… It’s mad… And nothing happens afterwards. They arrested some youth with hoodies, gloves and a pump-actions shotgun, and they got nothing. They took risks and the blokes laughed. And they pass their time wanting a coffee and a cigarette… At the moment, I see a lack of motivation; I expect a lot of the government. There are limits to prevention. (M. Gardib, gardien de la paix, 29, Sécurité Publique, 2002).
42 Among the repressives, there are very few who accord socioeconomic factors any influence: only 4.8 % of them cite the black economy and 5.6 %, poverty and unemployment. Moreover, of these police officers, only 1.7 % judge vigilance towards car drivers to be important, which is a much smaller proportion than in the other two groups.
43 As for the groups the police should target, they are above all criminal: drug dealers and users are the group against whom the repressives believe you should be most vigilant (for 29.5 % of them). Next come extremists (23.7 %), gangsters, pimps and prostitutes (19 %) and the group that is not a priori delinquent or criminal: immigrants. There are more supporters of repressive policing, compared to the median and preventive classes, who advocate a particular vigilance towards them (15.7 % of them), as this suggestion from a captain in the Sécurité Publique illustrates:
There is delinquency, and the Maghrebis are far from strangers to it. In terms of the GAV [gardes à vue] [watch list], it’s two-thirds Maghrebis. The reasons for this I’ll leave to others: the politicians. It’s not my problem. (M. Capam, captain, 50, Sécurité Publique).
45 Finally, one may wonder whether the weight of this group in the survey is not the result of the effect of the political context when the survey was carried out, a few months after the right has been returned to power in spring 2002. It is obviously difficult to measure such an effect. At most we can point out that the repressive ideology extolled by this group is the most common a year after the departure of the left who had launched Police de Proximité. We have called this type of police officer repressive because that have a more severe and closed conception of the role of the police.
Supporters of preventive policing: at the margins of the profession
46 The final group, the preventive group, representing 10 % of the survey sample, is by far the minority group. These are those that come closest to the Anglo-Saxon notion of “community policing.”  Certainly, like the medians and the repressives, the preventives believe that policing is primarily a profession that should ensure respect of the law (39.6 %), but much less overwhelmingly than the other two groups. Indeed, in contrast to the others, the preventives accord great importance to the protection of republican institutions (24.3 %). The care dimension in the “Anglo-Saxon” sense of the term comes in third position, while remaining greater than the other two categories: 18.5 % and 13.3 % respectively believe that emergency assistance and prevention of delinquency are priority missions for the police. We can see a picture emerge from the distribution of responses of a more “civil” conception of the profession, that is both more orientated towards rendering a service to the population and the state. In other words, the preventives opt in the majority for public service missions more than for pure repression.
47 These police officers are also the ones, and by some distance, who most highlight, socioeconomic factors of delinquency. 71.4 % of them attribute its rise to poverty and unemployment, against 10.3 % of the medians and 5.6 % of the repressives. Similarly, 11.3 % of them also cite the black economy among the causes of the rise of delinquency, something that is much less common among the repressives (4.8 %) and totally absent among the medians.
48 Presenting a mirror image of the repressives, very few preventives consider greater collaboration with magistrates to be useful (1.8 % against 46.7 % of the repressives) and a little fewer are inclined to promote partnerships with gendarmes (up to 29 % against 35.7 % of the medians and 45.9 % among the repressives). The fact that they accord greater importance than the repressives and the medians to the protection of republican institutions does not, however, imply that the solution to the problems of criminality are based for them uniquely on the regal functions of the state. Indeed, there are very markedly more preventives who advocate close collaboration with partners in civil society (educators, social workers, associations, doctors and teachers). 68.2 % of them nominated this close partnership as opposed to 5.7 % among the supporters of a repressive police.
49 Furthermore, more preventives also advocate vigilance towards extremists (30.2 % against 23.7 % for the repressives and 18.9 % for the medians). The preventives are, moreover, totally indifferent to the surveillance of immigrants (only 0.4 % of them are interested in this against 15.7 % of the repressives) and are not very inclined to accentuate the control of youth (4.6 % of them against 31.2 % of the medians and 10.5 % of the repressives).
50 The only point on which the preventives and the repressives converge concerns the vigilance towards gangsters (for 19 % of them) and drugs, which unites between 27 % and 30 % of police officers in each of the opinion classes.
51 From the three classes of police officers revealed by the latent class method we can establish a quite marked polarization between two extreme profiles, one very common one on the repressive side and the other a minority one on the preventive side, separated by a mass of police officers in an equivalent proportion to the repressive pole but with a much less clear-cut profile than the other two. This result invites a distinction between “the police” as an institution devoted to ensuring respect for social order and “police officers” (Robinson  1996) who hold more heterogeneous positions on how this same social order should be maintained. These diverse conceptions of the police, nevertheless, carry far from the same weight. As a result, we should put this pluralism of police opinions into perspective: few are the police officers who support a preventive conception of the police. Even though Monjardet has shown the volatility of police opinions in the past, and that the circumstances of the survey (Nicolas Sarkozy had just arrived at the Ministry of the Interior and he seems to have received votes in police milieus that were hostile to the Police de Proximité), we are compelled to observe that the mass have median or repressive opinions, situating the police doxa around a more or less high degree of repression, making the preventive position of police officers a marginal one.
An analysis, “all other things being equal,” of the distribution of the three opinion classes
52 Having described each opinion class and its respective weight in the ideological balance of power that courses through the police arena, it is important to understand their roots. The distribution of individuals in the three opinion classes was subsequently regressed to a set of professional and socio-demographic characteristics by means of a series of multinomial logistical models. 
53 This method consists of considering one of the categories of the dependent variable to be the reference category and estimating the effect of a series of independent variables on the probability of being located in the other response categories of the same variable. In our case the medians represent the reference category and we estimate the effect of a series of independent variables on the probability of being situated in each of the two others—here the repressives and preventives. Out of a concern to make our analysis exhaustive, the estimations presented in Table 3 do indeed reproduce all the possible contrasts: repressives/medians, preventives/medians, but also preventives/repressives. In the last column of each of the three parts of the table, the reference category is thus not the same as the two previous columns.
Logistical regression of the distribution within the three opinion classes according to various social and professional variables (Model 1)
|Repressives vs. Medians Preventives vs Medians Preventives vs Repressives|
|? p OR ? p OR ? p OR|
|Constant||0.084 -1.49 *** -1.574 ***|
Rank Senior officer|
Gardien de la paix
0.312 ** 1.4 0.259 -0.053
- 0.233 ** 0.8 0.077 0.311 * 1.4|
0.14 -1.256 -1.396
0.272 -0.082 -0.353
- 0.095 0.659 0.754|
0.182 0.298 0.116
0.051 0.072 0.021 - 0.132 0.142 0.273
Logistical regression of the distribution within the three opinion classes according to various social and professional variables (Model 1)Note: *** p < 0.001; ** p < 0.01; * p < 0.05.
|Repressives vs Medians Preventives vs Medians Preventives vs Repressives|
|? p OR ? p OR ? p OR|
|Constant||- 0.650 *** -1.376 *** -0.726 *|
Rank Senior officer|
Gardien de la paix
|- 0.351 ** 1.4 0.412 * 1.5 0.060 - 0.199 * 0.8 0.186 0.385 ** 1.5 - 0.265 -1.202 -1.467 * 0.2|
|- 0.256 0.033 -0.223 - 0.093 0.405 0.498 - 0.008 0.087 0.079 - 0.100 0.011 -0.088 - 0.096 0.166 0.263|
Most frequent Community policing
Examining evidence and clues
Contact with the public
Writing notes and reports
|- 0.121 0.123 0.002 - 0.196 * 1.2 -0.127 -0.323 * 0.7 - 0.190 -0.155 -0.346 - 0.782 *** 2.2 0.047 -0.735 * 0.5 - 0.229 * 1.3 -0.120 -0.349 * 0.7 - 0.463 *** 1.6 -0.294 -0.757 ** 0.5 - 0.414 *** 1.5 -0.146 -0.560 *** 0.6 - 0.157 -0.082 0.075 - 0.069 -0.347 -0.416 * 0.7 - 0.100 -0.144 -0.245 - 0.030 0.239 0.269 - 0.396 ** 1.5 0.282 -0.114 - 0.337 0.270 -0.067 - 0.303 * 1.4 0.336 0.033 - 0.042 -0.169 -0.211 - 0.273 * 1.3 -0.125 -0.398 * 0.7 - 0.174 -0.031 -0.205 - 0.143 -0.417 -0.560 * 0.6 - 0.621 -11.559 -10.939|
Dress Plain clothes|
|- 0.060 -0.190 -0.130|
|Satisfaction: 0 out of 3 disparity 1 out of 3 between 3 tasks 2 out of 3 and 3 skills 3 out of 3||
0.528 *** 1.7 0.213 -0.315|
0.297 ** 1.3 0.220 -0.076
0.152 -0.010 -0.163
|Repressives vs Medians Preventives vs Medians Preventives vs Repressives|
|? p OR ? p OR ? p OR|
|Constant||0.666 * -3.050 *** -3.716 ***|
Rank Senior officer|
Gardien de la paix
0.347 * 1.4 0.206 -0.141|
0.029 0.074 0.045
0.256 -1.517 * 0.2 -1.773 * 0.2
0.229 0.028 -0.202
- 0.099 0.252 0.352
- 0.007 -0.003 0.004|
0.101 0.052 -0.050 - 0.109 0.184 0.293
Most frequent Community policing
Examining evidence and clues
Contact with the public
Writing notes and reports
0.066 0.147 0.081|
0.091 -0.087 -0.178
0.128 -0.162 -0.290
0.700 ** 2.0 0.115 -0.585
0.176 -0.047 -0.223
0.431 ** 1.5 -0.351 -0.782 ** 0.5
0.357 *** 1.4 -0.148 -0.505 ** 0.6 - 0.176 -0.034 0.143
0.024 -0.316 -0.340
0.022 -0.109 -0.131 - 0.024 0.281 * 1.3 0.305 * 1.4
0.347 ** 1.4 0.307 -0.040
0.348 0.217 -0.131
0.305 * 1.4 0.352 0.047
0.069 -0.167 -0.236
0.330 ** 1.4 -0.197 -0.527 ** 0.6
0.102 0.043 -0.059
0.075 -0.409 -0.484 - 0.351 -12.727 -12.376
Dress Plain clothes|
|- 0.065 -0.145 -0.081|
|Satisfaction: 0 out of 3 disparity 1 out of 3 between 3 tasks 2 out of 3 and 3 skills 3 out of 3||
0.466 ** 1.6 0.250 -0.216|
0.307 ** 1.4 0.261 -0.046
0.156 * 1.2 0.006 -0.150
|- 0.208 * 0.8 -0.676 ** 0.5 -0.468 * 0.6|
|Age||- 0.026(a) *** 1.0(a) 0.016 * 1.0(a) 0.043 *** 1.0(a)|
|Repressives vs Medians Preventives vs Medians Preventives vs Repressives|
|? p OR ? p OR ? p OR|
Father’s Agricultural worker
occupation Shopkeeper, craftsperson,
Higher management, liberal profession
Higher ranked police or military officer
Lower ranked police or military officer
0.265 0.067 -0.199
- 0.159 0.014 0.173|
0.179 0.288 0.108
0.018 0.074 0.057
0.092 -0.097 -0.189
0.085 0.457 0.372
0.159 0.157 -0.003
Parents in the Neither
public sector Father|
Father and mother
|- 0.049 -0.153 -0.104 - 0.001 0.089 0.090 - 0.095 0.090 0.185|
Qualification Lower than bac|
Bac + 2
Bac + 3 and higher
- 0.078 0.026 0.103|
0.211 * 1.2 0.122 -0.088
0.246 0.353 0.107
|Place of residence Île-de-France during childhood Other||0.035 0.115 0.080|
|Occupation Police or military of spouse Other||0.150 -0.474 * 0.6 -0.624 ** 0.5|
Logistical regression of the distribution within the three opinion classes according to various social and professional variables (model 3)Note: *** p < 0.001; ** p < 0.01; * p < 0.05.
(a) Caution should be taken when reading the odds ratio calculated from continuous variable—here age—, for which the effect measured corresponds to the estimated variation of the relationship of chances associated with an increase in a unit—here a year—of the variable. The sign of the variation is more important than its amplitude which depends greatly on the unit of measurement of the variable.
The variables introduced into the models
54 The analysis carried out by means of logistical regression aims to determine to what extent the variations in the attitudes measured by the distribution in the three opinion classes, as defined by the latent class method, resist the control of these effects through a number of basic occupational characteristics. More precisely, here this consists of measuring the effects associated with rank and division. From this point of view, while the initial sample was made up principally of three corps, we have only included two. Commissaires [superintendents], the number of whom was too small, were excluded from the analysis, which is thus based on the senior officer corps (commanders, captains and lieutenants, who since the 1995 reform replaced inspectors and officiers de paix) and the supervisory and enforcement corps, within which we distinguished gardiens de la paix and sergeants, who constitute an intermediary rank between senior officers and the gardiens de la paix. As for the “others” category, this is a residual category that is difficult to interpret because it is heterogeneous, comprising adjoints de sécurité and administrators who incidentally completed the questionnaire. These will not by analysed as such.
55 The “division” corresponds to the departments, we include six in our sample: Sécurité Publique (SP), Police Judiciaire (PJ), Renseignements Généraux (RG), Compagnies Républicaines de Sécurité (CRS), Police Aux Frontières (PAF) and Direction de la Formation de la Police Nationale (DFPN). The analysis also makes a distinction between the police officers who carry out some or all of their work “in uniform” and those who work in plain clothes.
56 These first two variables were completed by the indication of the roles most frequently occupied by the police officers in the exercise of their duties. This information indicates, beyond the qualifications of the post, the function that is effectively most frequently fulfilled from a list of nineteen tasks (community policing, patrolling, guarding, surveillance, procedure, examining evidence and clues, questioning, interviewing witnesses, taking complaints, managing conflicts, contact with the public, maintaining order, arms drill, driving, management, giving orders, writing notes and reports, teaching/tutoring, other), which were introduced into the analysis as dichotomic variables while specifying, however, that respondents were constrained by the questionnaire to only specifying a maximum of three.
57 To these objective properties of activity was added a final characteristic that relates more to the perception that police officers have of the conditions under which they work, which we could assume would have an effect on their attitudes measured by the distribution within the three attitude profiles previously defined. This variable is constructed based on the combination of responses given to the question on the three most frequently occupied roles and on the responses to the question of the three roles the police officers declare to be the ones they are most accomplished at, chosen from the same categories as previously. The combination of responses given to these questions permits four configurations to be distinguished, which relate in some way to the four ordered categories on work satisfaction. The first category, which brings together 9.5 % of respondents, corresponds to the figure of police officers for whom there is no correspondence between the three missions they most frequently carry out and those they feel most accomplished at. The second category (26 % of respondents) corresponds to those for whom one of the three corresponds, the third (30 %) to those for whom two out of three match and the fourth (34.5 %) corresponds to a perfect match between their most frequent missions and those they perceive they are most accomplished at.
58 Finally, we added control variables based on situations prior to entering the profession. These control variables included sex, age, level of education (less than baccalaureate, baccalaureate, bac + 2, bac + 3 and higher) and socio-occupational category of father, filled in according to the nomenclature of occupations and socio-occupational categories (PCS) from INSEE (agricultural workers; shopkeepers, craftspeople, company directors; higher managers, liberal professionals; middle managers; employees, manual workers), but treating police officers and military personnel as two distinct separate categories according to rank (commissaires, inspectors and superior officers in the army and the police on the higher rank side and gardiens de la paix, sergeants, corporals and rank and file soldiers on the lower rank side). We also controlled for French region of origin—distinguishing between police officers who spent most of their childhood in Île-de-France from those who essentially grew up in the provinces; the influence of the occupation of spouse—distinguishing between police and military spouses and other spouses; as well as the effect associated with having parents who worked in the public services, a relatively common situation for the police officers interviewed in the survey, since this amounts to a little more than two-thirds of them.
59 In order to measure the influence of these variables more precisely three successive models were tested: in the first, only the variables of rank and division were included (Model 1). For the second, the variables of role occupied and work satisfaction as defined by the combination of missions carried out most frequently and the missions that the police officers felt most accomplished at, as well as uniformed work were added to the first two variables (Model 2); and, for the third, the socio-demographic characteristics of the survey respondent were added to all the previous variables in the previous models (Model 3). In evaluating the three models successively, we look to test the robustness of the effects associated with the “professional” variables introduced in the first model when they are controlled by the variables introduced in the two later models. The interpretation of the effect of the independent variables on the distribution of respondents to the survey in the three attitude classes is done in the same way as in the case of binary logistic regression. For the category variables, a reference position is designated; the effects of the other categories are then evaluated in relation to this. For the continuous variables, the evaluated parameter (?) measures the effect of a variation of a unit of the variable under consideration. The asterisks added to parameters correspond to the statistical significance of the effects. The table also shows the odds ratio (OR) for those parameters that differ significantly from zero. An odds ratio equal to 1 corresponds to the absence of an effect, an odds ratio above 1 indicates a positive effect which is greater when values are greater (increased relative probability) and an odds ratio that is less than one indicates a negative effect which is more pronounced as its value approaches 0 (decreased relative probability).
The preponderance of effects associated with professional position
60 The results of the analysis, “all other things being equal,” (Tables 3a, 3b and 3c) show, in the first instance, a link between rank, division and the distribution in the different opinion classes. Thus, the officers that make up the middle hierarchy are significantly distinguishable from the gardiens de la paix by a greater probability of being on the repressive side rather than in the median group.
61 This repressive tendency of officers withstands control by the introduction of the roles most frequently occupied (see Model 2), except that the attitudes of officers seem more polarized: even more repressive than median, they also seem significantly more preventive than median, but this second contrast seems less significant, and, above all it does not withstand control for the socio-demographic characteristics of police officers introduced in Model 3, in contrast to the repressive tendency. All other things being equal, moreover, the officers remain, according to the most developed model, appreciably and significantly more repressive than median relative to the gardiens de la paix. Is this because they are further from the ground than the gardiens de la paix, more specialized in criminal affairs and, as a result, more inclined to adopt a hard line, detached from the direct observation of more ordinary social relations? Is it also an effect of the judicial training of officers and the recent militarization of the officer corps?
62 The fact of being in uniform or not does not, on the other hand, seem to exert a significant effect on the ideological orientation of police officers. It seems as if the previously strong dividing line between police officers in uniform (lower ranks) and those out of uniform (higher ranks) has faded, notably since whether a uniform is worn or not depends less on rank than on the tasks exercised.
63 The case of sergeants merits special treatment: the latter indeed exhibit a greater probability than gardiens de la paix to be on the preventive rather than the repressive side, at least in the first two models that only take into account elements of professional positions. This rank, entirely made up of officers who have previously been gardiens de la paix promoted by way of internal examination (while half of more senior officers are recruited through open competition), is an intermediate position, between the officer hierarchy and the base, where they accumulate experience from on the ground which means they show solidarity with the gardiens de la paix and experience upward mobility, through continuous training, leading them to exercise a close leadership style. This dual position could result in a predisposition to a more preventive ideology. However, this apparent specificity of sergeants disappears when controlled for socio-demographic variables.
64 Division posting, whatever rank, exerts no significant effect on the distribution of attitudes in Model 1, which only takes into account professional position, which suggests that belonging to one police department rather than another does not significantly guide ideological attitudes. These results invalidate some preconceptions on the distribution of ideological positions in different categories of police officers; such as the idea that police officers posted to the Sécurité Publique or CRS divisions are distinguished by their more repressive attitudes. In the same way, while in the culture of the police, but also in the common sense, police officers in the Renseignements Généraux or in the training division are often perceived to be more reflective, “intellectuals” even, they do not appear here to be specifically associated with any of the three profiles once all the social and occupational characteristics are introduced into the analysis.
65 This absence of an effect of police division can be explained by the cross-disciplinary nature of some of the divisions, which tends to reduce their uniqueness. To take only two examples, maintenance of public order is no longer the monopoly of the CRS, since there are public order maintenance units in the Sécurité Publique and in the air and border police. In the same way the Police Judiciaire does not possess a monopoly over procedure, surveillance or examining evidence and clues. As a result, it is more pertinent to direct attention to the actual roles occupied, independent of rank, division posting or wearing of uniform, as is the case in Models 2 and 3. The results in this respect are more enlightening: the police missions carried out do actually have a very marked effect on the distribution in the three opinion classes.
66 Some tasks are thus more strongly associated with the repressive profile rather than with the median or preventive profiles. This is the case with the maintenance of public order, driving or arresting and patrolling tasks, but also with procedural, surveillance and examining evidence and clues tasks and with giving orders. This latter point again clarifies the attraction of more senior officers to the repressive profile: more than rank, it is the fact that they find themselves in a position to issue orders that distinguishes the repressives from other profiles.
67 Conversely, occupying a post in contact with the public is significantly associated, in the third model, with a greater inclination to the preventive profile than to the other two profiles. But this is the only task to show such an effect. Other quite similar tasks, such as community policing, taking complaints and managing conflicts do not seem to exert significant effects on the distribution between the three attitude profiles as defined by the classes of the model. In the same way, the so-called “intellectual” tasks in the profession, such as writing notes and reports, teaching and management exert no effect.
68 To summarize, professional ideologies seem to conform to a dual division of police work. A vertical division, on the one hand, between leadership tasks and those subordinated to them, and, on the other hand, a horizontal division between repressive and investigative work and other tasks. From this point of view, it is not surprising that the work of apprehending criminals on the street, of physical work (going on patrol, maintaining public order and making arrests) that are at the heart of the job,  incline towards repression. But the fact that this investigative work is also situated on the repressive rather than median or preventive side is a more unexpected result. The way a commissaire justifies, when interviewed, the “right-wing” leanings of investigative police officers undoubtedly clarifies this result:
It is true that we in the Police Judiciaire are more right-wing minded, in that on the left they defend the ideas of offenders rather than of the police. Thugs get the best roles; so this shocks us. (Mme Comidol, divisional commissaire, 49, PJ).
70 The effect associated with type of tasks exercised thus suggests that the professional socialization that develops through the exercise of a unique job reconfigures opinions, as if there is a reciprocal contaminative effect between certain police missions and the ideology that underpins them. In order to tolerate ensuring the maintenance of public order, it is beneficial not to consider prevention to be the primary mission of the police. This ideological conditioning by roles is so strong in this respect that it withstands control by socio-demographic variables (age, sex, social origin, education) for most of the previously cited tasks. Thus, giving orders, maintaining public order, making arrests and driving retain a positive affect on adherence to a repressive ideology and a repellent effect on adherence to a preventive ideology, while contact with the public places police officers on the preventive side.
71 Professional socialization thus seems to produce a self-fulfilling dynamic, a “squad effect,” which leads to embracing the dominant ideology of the assigned post and which, it should be emphasized, is especially true for the repressive ideology. Caution, however, should be shown in the interpretation of the direction of the relationship between professional situation and ideological orientation: do police officers adopt a repressive ideology because they occupy posts that involve them in repressive work or do they occupy these posts because they match their conception of the police? Only longitudinal observations and the gathering of qualitative data would allow us to determine this. Assignment to a particular post is only partially based on free choice, remaining in a post for a long time, in contrast, would undoubtedly be a gauge of satisfaction with a posting. Since we could not determine length of time spent in a post in the survey, we looked to approach this dimension by way of a measurement of professional satisfaction of police officers.
A repressive tendency and professional frustration
72 Police ideologies differ strongly and significantly according to a form of job satisfaction. This variable was constructed, let us recall, based on the degree of correlation between tasks most frequently carried out and those that police officers expressed they felt most accomplished at. A significant correlation becomes apparent in this respect: it is when there is a big gap between missions favoured by individuals and the roles assigned to them that the repressive tendency is most pronounced. It is possible to read in this a more general effect of professional frustration, with respect to which the attitudes of police officers are not so radically different from those observed in other areas of social life.
73 This effect, like the previous ones, is quite widely resistant to control for the socio-demographic characteristics of the police officers interviewed in the survey (Model 3). Thus, these effects do not stem from a selection effect tied to the social and cultural characteristics of the officers encountered in different ranks and charged with different types of mission but seems well and truly to result from the characteristics of socialization and professional experience within the world of the police.
74 Insofar as a substantial number of police officers, whatever their rank, do not get to choose their posts, in particular at the beginning of their careers, and find themselves subjected to substantial job rotation (with new ministerial policies, a new divisional head, transfers or promotions), professional frustration with not being able to totally match roles to skills is a quite common experience, favouring a repressive rather than a median ideology.
75 The consistent, quasi-linear effect of the job satisfaction variable invites us to reconsider the principle of versatility and interchangeability of police officers that leads to the perpetual reorganization of organizational structures, particularly in the Sécurité Publique (Lemaire 2008). This way of managing human resources, already underway when new management techniques were introduced (Matelly and Mouhanna 2007), contains a cost: the hardening of the ideologies of police officers.
The weak effects of social origin or education
76 In comparison to the effects associated with rank and the nature of tasks undertaken, the effects of characteristics introduced into the analysis as control variables seem relatively modest, despite the marked improvement in quality of fit to the data that comes from taking them into account, whatever the measure (pseudo R² or difference in likelihood ratios). The familial socialization characteristics, as captured by the father’s socio-occupational category (notably the fact of having a father in the police or military) and the employment status of parents (working in the public sector or not) in particular exerts no significant influence on the distribution in the three opinion classes.
77 Level of education does exert a significant and somewhat unexpected influence on the distribution, since it is practically only attached to those with a bac + 2 qualification having a tendency towards repression. Perhaps this result can also be explained by the same frustration process mentioned above, here associated with experiencing a relative decline in status which, in the police more than in other public bodies, is probably more concentrated in this group of graduates, for whom their social and professional resentment is expressed in an adherence to more repressive conceptions of policing (Peugny 2006). This effect, limited to a single level of education, contradicts the stereotype. We could indeed have imagined a correlation between higher levels of education and a preventive ideology. In a profession where the level of education required is lower in relationship to rank than in other public sectors, one might have expected that atypically overqualified police might show particular dispositions, at least to the preventive profile. This is not the case, here reinforcing the effect of professional position above social characteristics prior to entering the police.
78 The effects of age, sex and occupation of spouse, in contrast to those linked to social origin or education, are among the most pronounced. Significantly, one can firstly observe that attraction to the repressive profile reduces substantially with age, while it increases inversely for the preventive profile. Thus, the more police officers advance in years the greater their chances of being preventives rather than medians or repressives. That this variable affects each of the opinion classes in a coherent way shows its strength. In the absence of longitudinal data, it is, however, difficult to disentangle the effects of progressing through the life cycle from the effects associated with belonging to different cohorts, joining the profession in different contexts.
79 Interpretation of the age effect is made difficult by the fact that this is not, in reality, a characteristic that is independent of the exercise of the job itself. Thus, the effect of post held partially covers the pyramid of age: from the youngest doing the action jobs of the profession (making arrests, maintaining order) to the eldest doing the more peaceful jobs. The fact that the effects of age are resistant to being controlled for type of task fulfilled in a logistical model, encourages us to focus on the effect of advancing years as an effect of the experience acquired that distances police officers from the most oppressive attitudes that are also the more stereotypical and that are the most in tune with ideological representations forged when distanced from confronting the reality on the ground. What is more, this is the most usual interpretation of the profession: the older officers are more level-headed, more reflective and the young police officers are “cowboys” eager to be heroes in the fight against violence.
80 The hypothesis of a generation effect relates, in turn, more to the influence of recruitment contexts and career progress, which from the point of view of professional environment as well as from the point of view of security ideologies, very markedly differentiates generations of police. But the changes of political power between parties in the last twenty years makes such an approach difficult: in interviews, some anti-left police officers certainly emerged, heirs of the “Marcellin” or “Pasqua” eras, others were clearly loyal to the Socialist project, to Chevènement’s Police de Proximité, but these affiliations are not sufficient for there to be promotion of officers and gardiens de la paix of the same political complexion, as might have been the case for the rank of commissaire.
81 Measuring the influence of conceptions of the police promoted by such and such a government far exceeds the possibilities offered by the survey; we can, however, rely on the evolution of police recruitment: while in the 1970s and 1980s the competition to become an inspector was much less selective and permitted the absorption of a quite large variety of career paths, the creation of the officer competition in 1995 marked a clear break: the requirements to parade and wear uniform during training and when in position, the emphasis on judicial and leadership knowledge led to a whole section of candidates either to turn to the competition for commissaire or not to take the competitive exam, thus narrowing the profile of recruited officers. As testified by this captain, a tutor of student officers passing through his commissariat:
When there are student lieutenants, it’s me who takes them. I explain things to them, give them tips; they have to be channelled. I try to tell them to be more flexible. I had colleagues that wore flares and had long hair. Now I wouldn’t last more than a fortnight like that, I’d be fired. We’re training officers who are too militaristic. Those who come under me have no chance, because that is not what I advocate. (M. Capri, captain, 39, SP, 2002, recruited in 1982).
83 There is another major generational change, the level of education that police officers possess has increased more rapidly than the required level, notably in the rank of gardien de la paix, perhaps giving rise to a form of frustration among young recruits which, as we have seen, exerts an influence on ideological orientation, which can be partly understood by the effect of age.
84 This generational effect is finally corroborated by the characteristics of police recruitment as it can be shown in the survey (Pruvost and Roharik 2011). Indeed, we compared the two generations furthest apart in terms of age among the respondents (those recruited longest ago vs. the neophytes) in order to identify the socio-demographic contours of these two generations, but also the differences with respect to initial motivation, which are notably: the youngest report having chosen the police profession as a vocation for adventure rather than to carry out altruistic missions. Even if this is a retrospective recollection of initial motivations (biased by their current situation) and while it remains difficult to separate age and generational effects, two distinct types of vocation are clearly drawn that relate to two conceptions of the job.
85 Like the age effect, the effect of sex on a profession with barely any women (14 % in 2003) is difficult to analyse as an effect of a purely exogenous socio-demographic characteristic. As a minority in a profession where they are narrowly selected, women are clearly distinguished from men by a greater conformity with respect to the median profile and a more pronounced distance either from the repressive or preventive profiles. It seems that the statistically more improbable careers of women in the police are accompanied by a greater adherence to the professional doxa and a comparative distance from the extreme and “militant” postures. In other words, to facilitate their integration, women have every interest in distancing themselves and the figures of “social worker” or “fascist” (Pruvost 2007a).
86 The effect associated with spouse’s occupation also emphasizes the strictly professional context on the formation of values and opinions. Polices officers with spouses who are also police officers thus show a greater closeness to the repressive and median profiles in comparison to the preventive profile. This effect of occupation of spouse suggests that endogamy favours identification with repressive attitudes and a greater distance from preventive attitudes.
87 It should be noted that women plunged into a group that is masculine in the majority, have a greater chance of having a spouse in the police than vice versa. This is the case for 60 % of them. This confirms a point observed in the field of policing: endogamy, rather more than a police heritage (only 10 % of male or female police officers have a father in the profession), creates a clan-type solidarity that does not contribute to an opening up of the profession (Pruvost 2007a). The example of Mme Garlieuteb, reputed to have a strong authoritarian personality is symbolic: married to a CRS officer, she does not hesitate to make her marriage a gauge of her professional loyalty:
I couldn’t have married anyone else [outside of the police]. I live policing, eat policing, sleep policing. When my husband disrespects his colleagues, I remind him of the rule: “Once a CRS always a CRS.” (Mme Garlieuteb, 51, officer, SP).
89 Our survey shows that police ideologies, while plural, carry far from the same weight. Inductively, by the latent class method, we have certainly identified three positions, a preventive and a repressive conception of the police, and an intermediate position; but the analysis shows the strength of adherence to a professional doxa (Champy 2009) which seems to transcend the internal divisions in the world of the police and that it is hard to believe that this only reflects the specific contexts in which the data were collected (the 2002 presidential election, nomination of Nicolas Sarkozy as Minister of the Interior). The majority of police officers surveyed in 2003 lean towards a closed conception of the profession, which should focus on collaboration with the gendarmes-magistrates pairing (for the repressives) and the gendarmes-parents pairing (for the medians), and that delinquency is principally the result of the failings of the justice system (lack of sanctions, lack of prison places) for the repressives or lack of supervision by families or at school for the medians. The police officers that think that collaboration with partners in civil society and that delinquency is due to poverty and unemployment are in the minority.
90 While the attitudes of the medians do not lend themselves easily to interpretation in terms of political positioning, those of the repressives on the whole quite clearly echo the conceptions of policing supported in France by the right, while the positions expressed by the preventives bring more to mind those traditionally supported by the left. In short, although the survey does not provide direct information on political positions, even less on how police officers vote, our results substantiate a certain ideological closeness between the profession and the right, as quite radically expressed by a young officer, interviewed in 2001:
It is impossible to be police officer and integrate into a squad and hold forth on the left. If you’re not on the right and not racist, which is my case, it’s difficult, even in the BPM [Brigade de Protection des Mineurs] [child protection squad]. It’s difficult to try and justify integration. (M. Lieutef, lieutenant, 29, PJ, 2001).
92 The overriding influence of professional socialization of police officers on ideological positioning is another important result; so we might expect to see a pre-eminent effect of the “primary” socialization of police officers emerge, captured by the analysis of the socio-demographic characteristics prior to joining the police. Yet, the analysis reveals that a father’s education or socio-occupational status carries very little weight on the more or less, repressive, preventive or median orientations. As for the significant effects of sex (they make women lean towards the median side) and age (older people lean towards the preventive side), they relate back to methods of police recruitment and can also be interpreted as effects linked to professional socialization and the social morphology of the profession. While women are characterized by a quite neutral position, it is thus in reality in part because their minority status exerts a certain pressure for “discretion” and “self-restraint” in the explicit and implicit expression of their points of view (it is good for them not be noted for having extreme points of view). As for the effects of age, this probably, in part, captures that of accumulated experience, which is accompanied by a certain moderateness, as evidenced by a distancing, with age, from the repressive.
93 The effects of professional socialization appear more clearly still in light of the very significant effect of specifically professional variables: belonging to the rank of more senior officer or being assigned to certain tasks, such as patrolling, driving, judicial police work, maintenance of public order accentuate the probability of supporting a repressive conception of the police, revealing a coincidence between police ideologies and position held. The direction of these effects is however uncertain. Two hypotheses oppose each other: either police officers choose certain functions according to their ideological orientations, which is something we can not exclude but which is not strictly measurable, or opinions are moulded by police functions occupied.
94 These two options undoubtedly in part combine to be behind the “effects of position,” the study of which would require a combination of more detailed ethnographic studies (Fassin 2011), which would enable, for example, the effects of a squad in a commissariat to be observed, and the collection of longitudinal data, which would allow us to study the moulding of opinions over the course of a career. The demonstration in our survey of the weight of post currently held by officers in their conception of the profession could, from this point of view, explain why quantitative longitudinal studies carried out in the past identified a certain volatility in police opinions: changing posts frequently, police officers adopt an ideology that in some ways helps them make sense of their mission, in particular in the case of the repressive ideology (Monjardet and Gorgeon 1993, 1996, 1999, 2004).
95 Because, and this is another important result of our survey, there would appear to be no firm relationship between adherence to a preventive ideology and the occupation of posts that fall within “community policing” (community policing, teaching, managing conflicts, taking complaints), if it were not for the tasks that come under the banner of “contact with the public,” which seem to be substantially associated with the preventive profile rather than the other two (Model 3). One can undoubtedly interpret this result as a sign that the organization of the police does not promote a particular profile of police officer to these positions, neither does it develop an employment doctrine that is convincing enough to remould the opinions of its officers, who, in 2003, accord greater prestige to “action” operations than to the more invisible work of resolving conflicts peacefully.
|Primary mission of the police||Arrest delinquents||0.2177||0.0809||0.0579|
|Ensure respect for the law||0.5826||0.5700||0.4822|
|Rescue people in danger||0.1102||0.1351||0.1599|
|Protect republican institutions||0.0586||0.0923||0.1861|
|To reduce criminality, the police must collaborate more closely with||Parents||0.0256||0.2704||0.0553|
|Partners in civil society (doctors, associations, teachers)||0.0672||0.2693||0.5583|
|The increase in delinquency is primarily due to||Poverty and unemployment||0.0637||0.1303||0.4179|
|A lack of supervision in the family or at school||0.3299||0.7947||0.2846|
|A lack of sanctions in the justice system and prison places||0.5501||0.0665||0.2020|
|The black economy||0.0456||0.0001||0.0860|
|The fact there is too much crime in films and the media||0.0107||0.0084||0.0095|
|With respect to which categories do you think most vigilance should be shown?||Youth||0.1367||0.2853||0.1021|
|Gangsters, pimps and prostitution||0.1623||0.0740||0.1814|
|Drugs (drug traffickers, users)||0.2969||0.2967||0.2881|
Minister of the Interior in the government of the Socialist Prime Minister Lionel Jospin between 1997 and 2002.
Leader of the far right party, the Front National
These are the police officers who respond when you dial 17.
The “Honneur de la Police” was a far right terrorist group that claimed responsibility for assassinating Pierre Goldman in 1979.
The Socialist Prime Minister Lionel Jospin was defeated in the presidential election on the 21st April 2002 by Jean-Marie Le Pen, who qualified for the second round.
df = degrees of freedom.
Alain distinguishes reticents (use of force is acceptable, more ambivalent attitude towards derogatory behaviour, against its denunciation), relativists and conformists (more in favour of sanctions).
Monjardet and Gorgeon (2004), after the most recent carrying out of their longitudinal survey, noted an increase in one group, described as “centrist,” who do not take determined positions and had become the majority group.
The interviewees cited in the text are all anonymized by giving them fictional names.
Following Trojanowicz and Bucqueroux (1990), “community policing” can be described as a new philosophy of policing founded on the promotion of collaboration between police officers and citizens with the aim of regulating problems with criminality and public disorder, and problems within the home and between neighbours. This conception is very close to the one that in France inspired the so-called Police de Proximité policy at the end of the 1990s.
The method of regression is a known variant of binomial logistic regression that is a applied in cases where the dependent variable is a nominal variable with more than two categories and where these are not ordered, which is precisely the case with the three classes previously defined using the latent class method.
Let us recall that, while exercising various tasks, the police remain in fine “a hammer” (Monjardet 1996). For the debate between Brodeur and Bittner on whether physical coercion defines police work or not, see their respective articles in Déviance et Société in 2001.