Navigation – Plan du site
Articles ("Varia")

« Compstat » in Paris: a device for prompting police initiatives and accountability

« Compstat » à Paris : initiative et mise en responsabilité policière
Emmanuel Didier
Cet article est une traduction de :
« Compstat » à Paris : initiative et mise en responsabilité policière


“Compstat” is a neo-liberal quantitative management tool that has been invented in the NYPD during the 1990s and that has been imported in Paris during the 2000s. This ethnographic study shows how it works in France. It describes how the new use of numbers is associated to the high valuation of “prises d’initiatives” (discretion) and of “mise en responsabilité” (accountability) of the officers, the meaning of these two terms being closer to the domain of accounting than that of administrative law.

Haut de page

Notes de la rédaction

The original version (in French) of this article is available here:

Notes de l’auteur

Translator’s notes:

Certain terms were left in the original French:

Arrondissement = Administrative district in Paris. In New York City they use precinct. But a “precinct” is not the same as an arrondissement.

Commissaire = Captain, Commissioner or Superintendent.

Commissariat = Police station.

The Prefecture of Police of Paris, headed by the Prefect of Police, is an agency of the Government of France (and part of the French National Police) which provides the police force for the city of Paris and the surrounding three suburban departments of Hauts-de-Seine, Seine-Saint-Denis, and Val-de-Marne.

Texte intégral


1State reforms implemented in the last ten years have often been based on the idea that the State is too costly relative to the services it provides for its citizens; this often translates into the claim that the State’s agents are not doing enough. The result of this was the idea that public administration needed to borrow management models from the private sector because that sector knew how to motivate its employees and make them work harder. Sometimes the result was simple privatization of services-we are all familiar with the recent increased significance of private security companies (Ocqueteau, 2004) – other times, conforming to what some Anglo-Saxon authors have labelled The New Public Management (for example, Osborne, Gaebler, 1992), management techniques already used in capitalist enterprises were brought into the operations of the State.

2Here, we will present one such example. It relates to the use, since 2001, by the Prefecture of Police of Paris, of a management tool for public security policing called Compstat. This statistical tool was used to evaluate police commissaires and encourage them to maximize the number of police actions taken by their officers. Compstat had been implemented in New York City during the 1990s under the supervision of William Bratton (police commissioner under Mayor Rudolph Giuliani), who described himself as more of a CEO than a Commissioner. He also proudly claimed to have been inspired by the management theory of Hammer and Champy, called Reengineering (1993). This management theory was clearly created for the business world, where it achieved tremendous success (the cover of the French version trumpets “30,000 copies sold in France, 2 million worldwide!”). Michael Farrell, the head of the Department of Management Analysis and Planning for the New York City Police Department at the time of our fieldwork, who was also Bratton’s assistant and ghost writer during the 1990s, wrote at the time: we have adopted the Reengineering model of business enterprises (Farrell, 1994, 5).

3We would like to show how Compstat transformed one of the core functions of the State-that of public security policing, by encouraging activities called initiatives. This word comes from the realm of private management and has only recently penetrated the world of policing. In referring to similar but not identical police actions, we may recall that Philippe Robert (Robert et al., 1994) used the word proactive, Montjardet (1996) used selection of tasks and Renée Zauberman (1998) used decision - but none spoke of “initiatives” - as the police have for the last ten years.

4What we will show occurring on the ground is not insignificant-especially since the police world is not easy to access. Some people have deplored the “politics of numbers” to which the police have been subjected, or the “tyranny of numbers” that imposed itself upon the officers, but these analysts have seldom demonstrated, described or explained. Relying here on the full range of Alain Desrosière’s (1993, 2008) insights on the social effects of statistical analyses, we will demonstrate ethnographically how the Parisian version of Compstat functions, how the quantification of police activity allows the commissaires, and through reproduction over the entire hierarchical ladder, all police officers, to take initiatives. In other words, how, paradoxically, Compstat tallies these initiatives so as to contribute to their increase. And finally, we will show how these initiatives contribute to a generalized responsibilization of police officers. As Foucault explained in the late 1970s, this neo-liberal mechanism does not thus constrain agents of the public service, but rather encourages them to act more, to stimulate in them the exercise of their freedom (hence its name).

5Given the size of this article, our demonstration will involve at least two sacrifices. First, we will not describe the importation process of Compstat from New York to Paris nor explain why the Paris police department was so willing to adopt a technology that had been directly inspired by the private sector. This would amount to writing a history of public security policing from the point of view of its evaluation.

6Second, we will not present the criticisms targeting this method. On the contrary, we will adopt the point of view of the police officers who have agreed to speak to us or have allowed us to attend Compstat meetings, because this method seemed positive to them. They believed in Compstat and probably still do, and therefore saw no harm in explaining its mechanisms to us. Therefore, we won’t be asking any irritating questions (Laurens, 2007) but, on the contrary, will conduct comprehensive sociology, following the actors as they are in the process of shaping the world, following their will, their passion, their projects. Of course, there has been some criticism of the politics of numbers (in France, very little against Compstat specifically, because this technique is so little known, but the situation is completely different in New York) and this criticism describes what the State loses (and we can show that the State indeed does lose) when the State places police initiative at the top of its hierarchy of objectives (Didier, 2011). Yet, it won’t be of those who had hoped to move ahead, thanks to Compstat, that we will ask these crucial questions-these questions will be dealt with elsewhere.

7Since access to fieldwork in the police force is deemed to be difficult, we will first explain how our access was achieved. We will then describe how Compstat was put into place at the police prefecture as a mechanism to evaluate and motivate the central commissaires of the arrondissement. Finally, we will show that since it was implemented, Compstat equivalents and extensions have found their way to all levels of the police hierarchy, thus motivating all public security police to undertake initiatives.

I. Access to fieldwork

8Access to fieldwork was achieved step by step, from one contact to the next. It was first made possible by a commissaire I’d met in 2004 at the National Institute of Advanced Studies in Security (INHES) while I was working on issues of road safety. He had been a member of a mission sent to New York to observe Compstat and had become a believer in this tool. He agreed to explain its advantages to me in an interview. Following this discussion, the director of the INHES wrote a letter of recommendation for me addressed to the director of the Urban Community Police (DPUP) at the Prefecture of Police of Paris. The latter invited me to meet with the deputy-director of the territorial police who, after an interview, suggested that I continue with his deputy. At that point, I was hired by the CNRS, assigned to the Centre for Sociological Research on Law and Penal Institutions (CESDIP), which is a joint unit also financed by the Department of Justice which, I believe, increased my credibility in the eyes of the police and placed me in a very stimulating environment within which to study this issue. Over the course of several interviews, the assistant deputy-director of the territorial police explained the functioning of the Compstat statistics that he himself had generated.

9I told him that I wanted to see the effects of Compstat on the commissaire of a sector for whom the technique had been implemented. I asked to meet with a commissaire in charge of a sector (because the arrondissements had already been grouped into sectors) as well as the commissaire of the arrondissements. This was agreed to, but it was the administration that chose my interlocutors. It was suggested that I meet with a commissaire responsible for a sector and two commissaires from an arrondissement with different socio-demographic characteristics – one very posh and popular with tourists with the other being more working-class. Both had excellent statistical results.

10In one of the commissariats [police stations], I then asked to meet with the head of the Office of Operational Command (BCO).He is in charge of the major task of synthesizing statistics deriving from the different services and then forwarding them to the central commissaire. This was agreed to and we had several interviews and I visited its commissariat to observe the trajectory of the data. He let me conduct interviews with two of his assistants. In the other arrondissement, I requested and was allowed to meet with a lieutenant in charge of patrols. I was not allowed to meet with any patrol officers.

  • 1 Jean-Paul Proust was the Prefect of Police of Paris from 2001 to December 2004 and Pierre Mutz from (...)

11At that time, the Prefect of Police Jean-Paul Proust, was replaced by Pierre Mutz. I asked Mr. Proust for an interview, which he granted me before assuming his duties as Minister of State for Monaco1. At the same time, the sector commissaire that I had interviewed was appointed assistant to the deputy director of the territorial police. I asked him for a new interview because I knew that the evaluation of the central commissaires by Compstat occurred during very formal meetings, where they presented their results to the chief himself. I planned to take advantage of the situation to ask him to allow me to attend at least one meeting. He agreed and arranged for me to actually attend one of them. During this meeting, the new director of the urban community police noticed me and proposed granting me an interview, suggesting that he was looking for some ideas for improving these meetings. I think he also wanted to find out what I knew. The situation was a delicate one because he was asking me to act as an advisor, something which is not one of my usual tasks. However, I accepted in the hope-which was confirmed-that this would also bring me more information. During this entire period, I tried to get an interview with his predecessor, the former director of the PUP who had implemented Compstat, but that was not granted. These interviews, conducted with people who had promoted Compstat and who, for the most part, were proud of having implemented it, took place between April 2004 and December 2005.

12In 2009, I wanted to gather critiques of the method. I therefore asked some officials of the police union (high ranking officers) if I could meet with them. They agreed, but proved to be not very critical. On the contrary, even the representative of the National Union of Police Officers (SNOP), a union that had actually issued press releases against statistically-oriented evaluations, expressed positive opinions about Compstat in the interview. I therefore used this latest data here. In the end, I produced 20 semi-structured interviews and two participant observations.

II. The Compstat Variables

13In 1999, Jean-Pierre Chevènement, Minister of the Interior, implemented a thorough reform of the Prefecture of Police of Paris. He instituted the DPUP, which was meant to combat petty local crime, then on the rise. The neighborhood commissariats, which had hitherto depended upon the judicial police and the arrondissement commissariats which, in their turn, had depended upon public security, were placed under the sole authority of a central arrondissement commissaire. The reasoning was that local judicial inquiries and security-related actions would be coordinated and would re-enforce one another. Administratively, these forces were grouped within the sub-directorate of the territorial police, the second from the right in the DPUP flowchart presented below.

Figure 1: Flowchart of DPUP in 2003.

Figure 1: Flowchart of DPUP in 2003.

Our research covers the sub-division of the territorial police, that is the part of the organization chart beginning at the top and down to the three maps of Paris at the bottom.

  • 2 The deputy director of the territorial police explored the idea of audits conducted during visits t (...)

14Compstat was set up to evaluate these new central arrondissement commissaires. With a new position, a new evaluation mechanism was needed2. This one had the advantage of being based on a battery of statistics, a tool deemed capable of accurately assessing community police, whose mission is to process a multitude of small facts and, unlike the judicial police, rare but impressive “fancy cases”. It thus remains to be determined which statistics better describe this type of police.

15To present them, we can resort to the statistics which every commissaire studies in a select committee, with the deputy director of the territorial police or his assistant as well as their sector chief, on the eve of a solemn presentation to the Prefect. During such preparatory evaluation meetings, work is done on the most complete data. The figures are then in the form of an evaluation file, several centimeters thick, containing several folders: one includes all the information concerning the sector as a whole, and the others, having an almost identical structure, relate to each of the arrondissements individually. Each folder is summarized on a single sheet of paper where all attention is initially focused. This is an example of a summary that applies to a whole sector; the ones pertaining to the arrondissement commissariats are identical.

Figure 2: Example of a summary table discussed in a preparatory evaluation meeting.

Figure 2: Example of a summary table discussed in a preparatory evaluation meeting.

16The structure of this table was created by the Deputy Assistant Director of the territorial police after arbitration with his superiors. The prefect Proust also insisted that this structure had evolved, with the first tests, and the complaints of the commissaires who wanted certain variables to be added-those that seemed important to them (we’ll return to that issue) for understanding the constraints to which they had been subjected. In the end though, it was the conception of the territorial police as perceived by the high ranking officers that was adopted.

  • 3 The arrondissements themselves are divided into AIREs (Areas of Initiative, of Responsibility and o (...)

17This data is enclosed within an important spatio-temporal limit, since it forms the title of the table. The space can have sometimes been defined without difficulty, as it was for the arrondissements of Paris, or, on the contrary, after serious problems, as was the case with the sectors that did not differ enough from the arrondissement commissariats. Though at first too numerous to consider, they were reduced to three3. The chosen periods vary from several months to one year. They allow for chronological analysis in terms of evolution, symbolized by the arithmetic signs of + and – before the quantities, and reflect human actions.

  • 4 I gathered up all of my data in a working paper entitled Statistics for a new management of the pol (...)

We try to go over a period that means something for the people at the meeting [...] We try to motivate the personnel, placing superiors, and hence their collaborators, before the facts about the effectiveness or ineffectiveness of their action, the effectiveness or ineffectiveness of their mode of organization, compared to the results expected of them. So it’s a constant concern about efficiency. Going back a little too far, you would lose the substance. (Commissaire D., 2)4

18This spatial-temporal framework is the one that best allows the commissaires to be effective. Four major axes for defining the territorial police have been identified.

The territorial police] is deterrent-prevention, prevention relative to people and law enforcement actions when necessary, and God knows if it’s necessary, there is 80% of the work in this area, and then after [we must take into account] the means available to the deciders to accomplish all this. (Commissaire D. 61).

19The first three axes relate to activity. The deterrent-prevention refers to the street patrols, visibility, the creation of what is called PVMP “Points of Visibility and Meeting with the Public” (an important concept, but one that is difficult to rigorously define because of the vagueness associated with the idea of remaining visible (Ocqueteau, Pichon, 2008, 55-57). The prevention relative to people, or tertiary prevention, consists of meeting the merchants, the neighborhood residents, organizing interventions in schools. Finally, the law enforcement actions include the stops and arrests but also the attending to the calls, investigations and identification of suspects. The fourth axis pertains to the material resources-vehicles, weapons, vests, staffing, etc.-that determine the actions of actors beyond their mere involvement.

20From this axiom, a statistical "cognitive schematics" emerged, which took the form of a juxtaposition of ten boxes or tables on separate topics, each relating to one of the axes defined above.

21First, at the top left, we see the numbers which describe the forces available to the commissaire. Just to the right is the box which describes the overall activity of these forces, divided into judicial activity-it’s there that we find the crucial variables related to the solved cases, reported crimes and police log as well as administrative activity. Then, to the left of the activity, below the forces, is the availability of the different types of patrols to police the zones. School protocol consists of sending patrols to the exits at the busy times. The equipment is what allows or limits the policing activities. We notice the large difference between the numbers of vehicles that are theoretically and actually available, because accidents are frequent and repairs are slow (we will return to that issue). Then there is a first box pertaining to general criminality and a second box describing the main offences, the most sensitive ones. In this second box, we note two types of information, one below the other. First, we have the recorded cases, - conceived as an indicator of the number of reported incidents, and thus, in perspective, of committed acts. Second, we have arrest figures, an indicator of police activity, in response to these committed acts. It is indisputable that, from an external point of view, this report is nothing if not conventional, because not all of the acts committed within society are reported to the police (this is the whole issue of the dark figure of crime). But seen from the perspective of management, this convention is working, since it is indeed on the basis of reported incidents that the police must act. The other advantage of this box is that it is easy to change the title of its lines. At the time we conducted our interviews, robberies and burglaries were a priority, but this can change based on the priorities of the institution. This category thus gives flexibility to the system. Then comes the activity parameters of the neighbourhood police. They are measured in terms of trips to the crime scenes and preventive visits. The goal of this box is to show, on the one hand, what the officers are doing during the patrol (beyond the mere listing of patrols) and, on the other hand, the involvement of community policemen in judicial activity. Finally, we see the preventive actions and partnerships which, in this instance, consist of meetings with the residents or students. Those were put in place following the innovations of the Villepinte symposium of 1997.

22Finally, we note that there is no mention of what should not be done, of rules not to be broken, of cases where one should abstain. This portrait does not identify the number of blunders, of complaints to the Inspector General of services, or sanctions (Moreau de Bellaing, 2009). This portrait is about motivating activity. The rules limiting that activity are not considered.

23It should be noted that the figures used in this table do not all come from the same source. The data generated through the Policing Data Collecting Chart “4001” (crime and offenses) were provided by the Service for Prevention, Study and Anti-Crime Policy (SPEOAD) depicted on the chart above the three sectors. The data relating to the staff and materiel are provided by the Service for Management and Human Resources (SGORH), on the extreme right of the chart. Finally, the data relating to the activities of services, types of patrol availability, and prevention, come directly from sector central commissariats, and especially from their Office of Operational Command (BCO), which were created during the 1999 reform of the Prefecture of Police. Initially, the function of these offices was to harmonize the activities of the various commissariat services (mainly the recording of complaints and the local services of the judicial police).

24It seemed that this transformation required intense efforts to record everyone’s activities, which in turn greatly facilitated the work of quantification. For Compstat, all that was needed was to request that the BCOs transmit their statistical data to the sub-directorate of the territorial police rather than keeping them at the central commissariat. The number of patrols made during the month, the evolution of staff in the investigative divisions, or else the arrests distributed among the units of each sector, are three noteworthy examples of data monitored by the central commissaires who expressed interest (they were labelled the good students by the assistant deputy-director)-but this data was not systematically synthesized before Compstat.

25Thus, the data used by Compstat was available before Compstat was implemented, but the data was produced by different services and scattered throughout the organization. Therefore, the innovation consisted mainly in the centralization of the data and its being made available to the central commissaire-the only person in his commissariat having such a picture of his sector-and his hierarchy. Compstat was, therefore, not a “transformation into statistics” of the local police, but rather, a re-organization and a new use of figures that were already available.

III. What is the initiative of a commissaire?

26How is this data used? To answer this question, we must understand that they are not individually original, but they become so when combined, side by side in those ten boxes. Our approach, therefore, does not consist of studying the presuppositions, histories and effects of using one or another variable individually. This is a very interesting issue that has already been considered-the 4001, for example, is the object of a rich literature among which we can identify excellence (Aubusson de Cavarlay, 1996). Our approach is rather to understand the presuppositions and effects of their grouping together, their rapprochement. This aggregation takes precedence over each variable because, taken individually, they become interchangeable, as the boxes for “main offences” show with particular clarity. In this context, we can note that it is first the commissaires themselves who use this data; they do so in four different ways which constitute what they define as taking initiatives.

27The first can be called a « comparative reading ». It involves making connections among different boxes so as to develop a causal analysis of crime. Commissaire D. explains the subtleties of this using the example of burglaries (in the box main offences, recorded cases). They were decreasing and the demonstration would be more persuasive if this had been the opposite, but this was not the case. So, for the sake of the exercise, he tries to understand what could account for this decrease. He then refers to patrols in the compartment types of patrols available and sees that they were decreasing as well. He infers that is not this method of deterrence that explains the evolution of crime. The explanation must lie elsewhere, he suggests-perhaps in the wide adoption of some technical device, such as better locks. The important thing in this analysis is that the figures allowed him to link criminal activities with police activity. The two boxes, once brought into relation with each other, allow one to measure the degree to which police activity can influence criminality. However, there is no causal link between one and the other. Highlighting this relationship simply gives the commissaire the chance to estimate the impact of his initiatives.

This is a correlation… more or less fictive, to be honest. Because I do not believe that burglaries depend on police activity. [...] But it is a step that allows for the responsibilization of everyone, regarding what does not work or what works well in a sector. It is for each individual to make up for the 10, 20, 30, 40% I do not know, nobody will ever know what impact someone can have on this issue, if one gets better (Commissaire D., 70).

28One can ask oneself whether or not there is a self-referential circle here. Indeed, it is the police themselves who record the complaints regarding burglaries, and then it is also they who record whether the complaints have been resolved or not. Is there no place then for a congenial arrangement of the results? However, one would point out that there are no other statistics. Indeed, the only non-police data that could eventually play the same role might be the victimization surveys, but conducting them at a steady pace for such small areas would require an investment that is clearly unsustainable (Robert et al., 2008).

29Following that, the second use of this data, the commissaires are expected to react to a problem:

[Before the meeting, the commissaires] prepare the responses. When something doesn’t work, they indicate that they saw something was wrong and they say “here are the decisions that I have made, and they should bear fruit in about two to three weeks”. This allows them, in the absence of having a logical explanation, to at least be able to say “I took action”, “I took such a measure in order to straighten things out”. The worst thing would be “you see, this isn’t good”. “Yes”. “What will you do?” « I don’t know ». That’s catastrophic. Now we say: « well, here we are, we have a marked phenomenon, we had a gang that came and targeted the basements for a while, so we had an 80% increase in burglaries”. “Okay”. “What are you going to do?” « Well, I contacted the landlords, we toured the basements, we re-enforced all the exits, etc. The patrols will follow-up for several nights”. (Commissaire D., 60)

30We note that this response is not considered to be already in the statistical tables. The police, unlike some social science researchers who specialize in “affordance” (Gibson, 1977), do not endow their tables with their own imaginations. The response to a problem inevitably comes from the commissaire who proposes something on his own initiative.

31Third use of the tables, once problems have been identified and responses executed: the tables facilitate measuring the success of those responses. A commissaire, head of the sector, explained this point during an interview. He was discussing the “eternal” problem of street peddlers in the heart of one of the main tourist sites in Paris:

They arrive in the morning at 9:30–10:00. And, depending on the season, stay until 10:00 PM or even 1:00 in the morning. In the morning, at 9:30–10:00, it is Punjabis especially that arrive. Then, a bit later, they are joined by Malians and Senegalese, and then a bit later, people from the Maghreb, and then finally, Chinese drink vendors. Okay, all this is not simple and it is already a first element of complexity (Interview with head of 1st sector, 92).

32Other complications connected to street peddlers: firstly, this offence is sanctioned with a ticket, so the punishments are limited (no incarceration) and secondly, those who are illegals are nationals from countries that don’t often facilitate the return of these people (especially India). It is therefore very difficult to have a long term impact against offenders. The commissaire explained, however, that after careful consideration, a large operation was implemented in March.

Concretely, there is an operation that is initially conducted by policemen in plain clothes who establish themselves in the area. They are then joined by policemen in uniform, who secure the area. And so a net is created and we catch all who can be caught. (Ibid.)

33The commissaire then analyzes the effectiveness of this operation in almost purely quantitative terms: two hundred police officers were involved in the operation, 107 foreign nationals who were peddling were arrested, 48 were taken to the RG because of their irregular status. Afterwards, he notes that the number of tickets for street peddling fell drastically, even if, unfortunately, it is currently rising (“104 in May, 142 in June, 131 in July”), a rise which he explains by the fact that the offenders have gradually returned to their areas of activity. And so we are able to determine, because of the quantitative data, that the operation was a success, albeit a short-lived one.

34So, the data that is used to facilitate initiatives is also used, after a while, to evaluate those initiatives. This is the same table used to act and to assess, the only difference being that the dates are not precisely the same. We thus reach the surprising conclusion that the evaluation directly contributes to the action because analysis of the following action will relate to the same table which was used to evaluate the preceding initiative, obviously with the only possible conclusions.

35Fourth use outlined above, the tables are also used to highlight the limits that materiel can impose upon the actions of the commissaire (even if at the Prefecture of Police of Paris, the officials tend to congratulate themselves for the scale of materiel means at their disposal rather than their deficiencies). For example, the commissaires derived, from the chief of police, the fact that beyond the mere possession of vehicles by the commissariat, it was necessary to show their availability in the tables, because the slowness of repairs makes the two amounts very different. To the extent that Compstat is used to show the initiatives taken by the commissaires, it is exactly within the purpose of the tool and of those who have implemented it, to show the material obstacles to their actions.

36Finally, the tables are also an opportunity to discuss, more broadly, the general situation of the police in its arrondissement:

Beyond examining the retro-projection, [we can] consider, somewhat, the not yet quantified problems of the arrondissement. Here we come to the verbal aspect, the absolutely essential complement, beyond the solely quantified retro-projection-it is the session wherein we discuss the problems we encounter. What are these problems? When you have the commissaires of V, or C, or of D, it is obvious that you are dealing with a certain number of inner cities or sites that are more sensitive and where disorder was always an important issue. How are things? What is the current relationship with the police? We have had incidents of police or fire vehicles being stoned, incidents of urban violence, depending on the arrondissements, the incidents of street peddling, each has its own point of view. (Commissaire D.61)

37The quantitative variables and the numbers are thus a preparatory stage (« beyond the retro-projection ») to the creation of a synthesis regarding the general state of an arrondissement. There is no opposition here between the qualitative and the quantitative. On the contrary, one leads to the other and vice versa, with no discontinuities.

38The head of the sector also assists the pre-assessment session because, as we have seen, some phenomena extend beyond the limits of a single arrondissement. Aside from the fact that street peddling around a tourist site can span two arrondissements, there was, during the time I was in the field, a significant increase in the number of “Italian-style thefts” (two people on a scooter, one driving and the other snatching hand-bags) throughout a sector. It was more than likely due to a single team going around a whole area of Paris. A successful response required that action be taken at a more global level. This set of boxes can thus be read at two different scales: that of the arrondissement and that of the sector.

  • 5 The police initiative as conceived here is very close to that which the Americans call « police dis (...)

39So, faced with the difficulty of precisely identifying the actions of police commissaires related to public security, the pre-assessment sessions encouraged them to take initiatives. A highly complex statistical mechanism allows them to personally identify the flows of crime that are connected to the activity parameters of their officers, then to create a response to a problem deriving from these variations. At the same time, the statistical table allows them to evaluate the quality of responses they provided in the past and thus to eventually improve them. The framework provided by the Compstat tables thus leads to what the Anglo Saxons call an empowerment, that is to say, it gives the commissaire the power to act individually against developing disorders5. The statistics form a kind of background against which the action of the commissaire stands out in relief. All the actions that the commissaires can attribute to a single individual are labelled as initiatives.

40In other terms, Compstat almost creates the famous “culture of the result” with which it is often associated, except that, although it may effectively provide the means to highlight the results obtained by everyone within the material limitations which constrain them, the assistant to the sub-director of the territorial police admits that it is still difficult to generalize this to a whole... culture (“the term ‘culture’ is a bit presumptuous” D.63). Compstat thus clearly fits within a neo-liberal logic as defined by Foucault (2004b)), which places each commissaire, like a business leader facing his shareholders, in the situation of having to prove his own added value to the struggle against disorder. Proof of which he will now have to provide in person to the chief of police.

IV. The Plenary Session and the Responsibilization

  • 6 Symptomatically, the public that had initially been accepted in New York was replaced in France by. (...)

41The conclusions derived from the pre-evaluation sessions are then reiterated in more formal evaluation meetings. These meetings are held every Thursday morning, between 9:00AM and 11:00AM. During my fieldwork, two weeks per month were reserved for testing the mechanism in services other than territorial police (the judicial police, transport police, control of public order and traffic and general information, alternately). These meetings were apparently not fruitful, so that in the end, the central arrondissement commissaires continued to show up there about once every two months. Although they are absolutely not public, I was able to attend on November 24, 20056. We can also see in the meeting plan provided to people in attendance (figure 3), that no place was really prepared for me. How are the commissaires evaluated? We will explain that by proceeding to an ethnographic description of what is nothing less than a contemporary statistical ritual.

1) The Gathering

42In New York, the meetings are held at 6:00AM, in Paris, at 9:00AM. This illustrates the paradigmatic changes that Compstat was subjected to in crossing the Atlantic-in general, it was made softer and less threatening in terms of the procedure and sanctions involved.

43Fifteen minutes before the start of the meeting, the commissaires and their department heads, in uniform, are in a room in the prefecture. There are chairs, tables, and coffee, and everyone-around 15 people-stands around chatting, fairly relaxed. Ten minutes before 9:00AM, they are joined by the deputy director of the territorial police and his assistant, and a few minutes later, we hear a thunderous attention! which precedes the entrance of the director of the PUP himself, his assistant and his other deputy-directors. The contrast is striking because these new arrivals are not in uniform, like the commissaires, but rather in white shirts with rolled-up sleeves. The latter, whose offices are in the Prefecture of Police, are “at home” and look like it, while everyone else comes from the outside, so are formally dressed. The director of the PUP briefly takes the floor to convey his main message. In this case, he says that it is a good week, that there won’t be any problems, but he does not like the claim that there is a lack of manpower because he has just acquired a whole group of officers from the judicial police.

  • 7 Bad luck, I wasn’t aware of current practices and am one of the first to shake his hand: Hello sir, (...)

44Then, everyone leaves for the « hall of the 167 police officers who died for the Liberation of Paris », which requires us to cross the courtyard of the prefecture in a bit of muddled mob, and then descend a staircase into the basement. Again, some coffee is served by some police officers in uniform. There are some casual conversations for a few minutes; the chief of police then arrives with the main members of his cabinet. After greetings are exchanged7, everyone takes a seat (here is the seating plan that was distributed during the meeting):

Figure 3: Seating plan for the evaluation meeting

Figure 3: Seating plan for the evaluation meeting
  • 8 And it is difficult to describe the extent to which the sociologist, intimidated and hiding inside (...)

45It is impossible not to be impressed by such an assembly. There are many people (around 40), and they are not just police officers and thus invested with special powers; but beyond that, for the most part, very highly situated in the police hierarchy. It is very clear that we are within a group of powerful people, and these people do not deny themselves the pleasures of being so8. They therefore had me placed next to the only “technician”, who would be in charge of scrolling through the slides during the presentation. Together, we formed the rank and file. However, the mere fact that there is a table and some coffee is congruent with the “friendly” approach to the meeting desired by the prefect Proust, an approach which is contrary to the aggressive version observed in New York, where the police officers were called to the stand as though they were at a trial. Proust insisted: in order for the process to be accepted, it had to be deemed pleasant by the troops.

46The officials are grouped according to commissariat: on the right side, at the end of the table, are the heads of an arrondissement and, on the left side, one after another, those from two other arrondissements. To my right are the deputy directors of the DPUP and the head of the sector. Finally, at the head of the table, the prefect of police himself, his closest associates and the director of the PUP.

Figure 4: Overview of a meeting (an extract from Liaison, No. 82, 2002).

Figure 4: Overview of a meeting (an extract from Liaison, No. 82, 2002).
  • 9 Liaison is the internal bulletin of the headquarters of the Paris police.

During the meeting that I attended, the technician projecting the images was seated at the table and not in the middle of the room9.

47The prefect then makes a brief opening statement of a very general nature in a cordial tone. The first remarks are positive (in this particular case, it was just after the riots of 2005 and he congratulates the troops for their behavior); then he mentions a few problems (the first among them was that these troops had not reached the targets set by the minister for the expulsion of illegal aliens).

48Finally, he invites the deputy-director of the territorial police to begin. The deputy director stands up and begins his commentary on the slide show. The slides were prepared by four police officers stationed in SPEOAD-so states the complete evaluation report in an abbreviated form. First, the statistics for the entire sector, which would not be raised at that time, (but they could be at another time, I was told), then the statistics-arrondissement by arrondissement. During the interview, Proust insisted that he was relying on the deputy director to remain neutral, to present the results “without value judgements”. During the meeting, I noted that he did indeed adopt the manner of a master of ceremonies. His task, therefore, was to avoid conflicts of interest, to give voice to those statistics deemed to be objective, neutral and disinterested.

2) The Presentation of Statistics

49The slides are presented by arrondissement according to a strict order, to allow for a better understanding of the meeting: first a synthesis, then the criminality represented by the 4001 aggregates, then the activity parameters related to the facts (crimes solved and initiatives) and to the people (indictment, those in custody, deferments), a presentation on the activity of emergency police, a presentation on partnership activities, and finally, a presentation on the changes in human resources

50There are 21 slides for the first sector and 22 related to arrondissements-for a total of 80 shown in less than two hours; with time devoted to discussions, we can estimate one slide per minute. The presentation proceeds at such a speed that, without having studied the statistics beforehand, one cannot have actually understood what was going on nor been capable of noticing anything that would not have been orally explained.

51That is why a synthetic form was introduced: the colored arrow, which emphasizes change. This is an element that we didn’t focus on while studying the summary sheet during the pre-meeting evaluation, but which here is of major importance: the changes presented to the prefect are all highlighted. During the pre-meeting evaluation, (see top of the sheet) they were positive, stable, or negative, depending on their meaning; during the meeting itself, they were colored red (negative) or green (positive). Hence, the vital importance of the red arrow.

All of this effort is reflected in the system which the deputy-director holds dear, the distribution of arrows, green and red, showing at a glance what is important. The eyes of the high ranking officials are, of course, much more focused on the red arrows than the green ones (Commissaire D., 68).

52During his intervention, the red arrows, which ascend or descend, depending on the meaning of the variable, are referred to as a problem by the deputy-director, along with other characteristics of the sector or arrondissement.

Figure 5: The first three slides from the presentation I attended.

Figure 5: The first three slides from the presentation I attended.

The first one is the summary of the sector as a whole, and we immediately see (if we have color) that the arrow relating to the third variable “violent crime against people” is red.

53At the end of the presentation relating to his arrondissement, the principal commissaire spoke about certain things that were identified during the pre-eval13uation meeting, things that he thought were important but which were not always connected to what we had just seen in the slides. Two of the commissaires focused on the difficulties related to tourist sites and retail areas as the Christmas holidays were approaching, while the third commissaire mentioned nuisances connected to the sale of alcohol late at night. The commissaires were impressively eloquent and concise. Each spoke briefly and very effectively.

54These speaking engagements were not conducted just for the prefect, but also for the benefit of colleagues from other arrondissements, who acted as the audience. One of the commissaires, who was preparing to retire, was doing his best to make his colleagues smile, with a manner that was not so much irreverent as fun-loving. You could almost say that he was clowning around... but that might be going too far. These meetings were thus also situations where the commissaires were evaluating each other, judging their respective performances, and taking those judgements into account.

  • 10 Here it may be necessary to provide some subtext to this question. In France, one can normally lodg (...)

55It was only after this double presentation – the slide show and then the intervention of the commissaire – that the prefect or a member of his office spoke. In the meeting that I attended, the director of the PUP asked the first commissaire, from a very touristic arrondissement, to specify the percentage of crimes that had been reported but not committed in his arrondissement; the answer: 34%10. The prefect asked a second commissaire if it was really necessary to fight against the unlicensed sale of chestnuts in front of department stores (the holiday season was approaching). Indeed, it seemed to him that this sale of chestnuts had almost become a tradition-to the point where he had almost brought chestnuts with him to the meeting. The central commissaire answered that if he tolerated that activity, then he would no longer know where to place limits, especially since it was a Pakistani gang, and that in the end, fire was involved. The prefect did not answer. Finally, the prefect’s chief of staff asked the third commissaire if the order issued by the prefect allowing him to forbid the sale of alcohol at night (which caused a lot of public disorder in his arrondissement) had been intentionally ignored. The commissaire answered: yes, it is being ignored-for the man returning home late from work and buying a camembert and a litre of wine for supper-but no, it is never ignored for the pathetic losers who disturb the peace. The prefect joked that in honor of his retirement, he was going to ignore the two red arrows in his slide show.

56However, there were almost no red arrows during the whole meeting, so it was fairly quiet. “There was not very much fun. It’s more fun when there are more red arrows. Then you can really do some thinking” (Observation 1, 10) said one commissaire, upon exiting. So, the analysis relates to a large number of different variables meant to describe all the police activity, and the goal is simply to be positive about all these dimensions, that is, to not have any red arrows or not many, and if so, to be able to explain them.

57At the end of this meeting, the prefect concluded quickly and quite warmly, and offered his services to his troops: if you need prefectural orders, do not hesitate. One of the possible initiatives available to the commissaires is asking the prefect to take measures in the form of decrees, that is, to adapt the laws to local situations. To produce green quantitative variations, it is possible to mobilize decrees, and thus law. The law can thus be at the service of management, which involves producing quantitative variations in the right direction.

58Finally, the meeting was adjourned at 11:00AM on the dot. The prefect and his staff left very quickly. However, the hierarchy of the DPUP remained to discuss things a bit more. In sum, the general ambiance of the meeting was not as stressful for the police officers “put on the spot”-as it was for the New Yorkers in the same years. Proust wanted to transform the American meetings that resembled a trial or an execution (Commissaire Q., 49) into something more collegial, allowing for more informal exchanges among the commissaires themselves and between the prefect and his commissaires. There are many exchanges among colleagues, and the fact that the prefect offered his services at the end of the meeting can be interpreted as an additional desire to reduce the asymmetries created by hierarchy. However, we detect in it both an implicit comparison among peers in the presence of the chief and a method for the evaluation of commissaires’ behaviors (even if it involves an offer of assistance).

59The plenary session thus allowed the commissaires to present to the prefect and discuss with him the initiatives they were taking to defeat crime. At the same time, by means of the simple and effective colored arrow, the session also served to evaluate their performance. This crucial element, the positive or negative sanction of actions, is a concrete realization of a paradox often expressed in interviews: the fact that the commissaires are “responsibilized”. They very much remain subject to the hierarchical dependence upon their prefect, who sanctions them (the commissaires are “subjected to Compstat”, if we might phrase it that way-that is to say, they abide by this mechanism, which will ultimately serve to sanction them). However, this mechanism is organized in such a way that the individual does not submit but rather acts. The individual is not simply one who takes initiatives as an autonomous entrepreneur, he is also and simultaneously one who subjects himself to an evaluation by which he will be either tarred with the brush of blame or honored with the laurels of success.

60This situation of responsibility is better defined by the term accountability, which is business-related, rather than the more legal-oriented term responsibility. Indeed, at this stage, law certainly enters into the composition of the mechanism, for example, that is what defines the titles of most of the columns in the statistical tables. But law, as it occurs here, appears to be a series of opportunities for action, rather than a limitation. The police officer has the right to take a person into custody, he has the right to write up reports, etc. The example of the decree, which was proposed by the prefect, is paradigmatic in this respect: it accompanies the action rather than restrains it. The responsibility in question here is not aimed at attributing fault, but rather more at positively or negatively evaluating a series of actions. It is thus neither a question of legal or administrative responsibility (Dubois, 1996). This responsibilization is a managerial evaluation. Its advantage is that it thus makes submission to authority and autonomous action mutually compatible, since it is conceived as a responsible initiative.

  • 11 E. Hache notes many examples from the feminist literature that demonstrate just how responsibilizat (...)

61However, this individualization of initiative and distribution of responsibility might seem highly suspect to any sociological mind. Thus, in an excellent article on responsibility, Émilie Hache (Hache, 2007) not only doubts that we can individualize action at this point, but also sees in this process a new “exploitation principle”. She begins by highlighting individual autonomy that is associated with responsibility and so demonstrates the generality of processes that we have observed with Compstat. Then she points out that « society is mainly based on relationships of dependence ». Indeed, when we consider the crowd of human beings (everyone around the table and every unseen person who participated in the production of the statistics) and the mountains of artefacts (tables, cars, computers, bullet-proof vests, law, decree, etc.) necessary to produce the mechanism by which a commissaire could finally say I took an initiative for which I will be individually evaluated - one can only agree with her. Yet this remark leads her to conclude that the responsibility and the autonomy of these mechanisms are in fact nothing but falsehoods, that they “only make invisible” the dependence relationships, so that some people benefit from these connections without paying the price. The responsibilization would only be a “new sexual and social division among individuals capable or not, of responsible behavior”11. So what is it? Exploitation, or not?

V. Along the chain of command

62If Compstat creates an exploitation, it is that it creates an unequal distribution of a good. It seems to us that the first goods that can be identified, from Compstat’s perspective, are the margins of initiative left in the work, which are always valued. It could be that the mechanism gives a lot of freedom to some (the central commissaires) at the cost of a greater submission by their subordinates. It would thus not multiply initiative-on the contrary, it would unfairly reserve them for an elite. To learn whether this is the case, we will move along the chain of command and observe the effects that Compstat has had on every level of work.

  • 12 The creation of a « Club of Prefects » (Le Monde, October 13, 2009), p. 12) seems indeed motivated (...)

63We can begin our observation at the highest levels of the administrative hierarchy. Let us recall that for several years, the press has readily reported that the highest ranking people employed by the state have their performances evaluated statistically, in meetings that we have not had personal access to, but which seem to resemble those taking place in Compstat. This is the case with ministers themselves (Pezet, Ponem, 2008) - and so it is in our case - with the Minister of the Interior - and this also seems to be the case with prefects12.

64Closer to our fieldwork, every Tuesday morning, the deputy director of the territorial police organizes a meeting with the heads of public safety for the three sectors, that is to say, with the next highest level above the central commissaires of the sectors (cf. Figure 1). During this meeting, the main activity variables are discussed. After, they organize ipso facto, Tuesday afternoon, similar meetings with their central commissaires.

The Tuesday afternoon meetings are for the head of the sector, such as myself, the opportunity to comment on the crime observed in the arrondissements, the main recent manifestations, I have the latest figures, I comment, I sometimes ask the arrondissements to explain the negative developments. [...] It’s a chance to repeat what was said Tuesday morning at Mr. Q’s. [the deputy director of the territorial police], thus to pass along the message of management, and then at the same time, to also pass along my messages about one thing or another that I feel is important. (Commissaire, head of the first sector, 97)

65These meetings are supported by a table of statistics that is quite comparable to, but a significantly simplified version of, the sector synthesis presented to the prefect. They serve to analyze the variations of the principal clusters of crime and activity as measured by the 4001 record (line) reported to the various arrondissements (column). These variations are framed in temporal terms (one year is compared with another) as well as in geographical terms (the arrondissements are compared with each other). The analysis can be made in any direction, but mainly by comparing the “Evolution” columns of different arrondissements.

Figure 6 : Tableau d’évaluation Secteur le 15 novembre 2004.

Figure 6 : Tableau d’évaluation Secteur le 15 novembre 2004.

66What counts for each central commissaire, is to show that thanks to his initiatives, he achieved solid progress, and if possible, as solid as those achieved by other arrondissements. The mechanism is thus comparable to that of Compstat, used at the prefecture, with the comparison here between teams significantly strengthened by the contiguity of columns. The central commissaires are not spared. As soon as they return to their arrondissements, they get together with their subordinates and compare their statistical results. A central commissaire explains:

Later, it is very clear that when I return with my file from the evaluation meeting, the first thing that I do is a debriefing with the intermediate hierarchy. This does not serve to motivate and responsibilize me. From the moment I am responsibilized and motivated, I have to responsibilize and motivate the others. Down the chain of command-that’s logical. Because it could only be me that was responsible for the arrondissements’ results. I’m the one who works the machine, but it’s the frontline that provides the work. [...] And when I have a meeting with the heads of the brigades, they get the document from me. I also give them a document on the crime trends. The monthly crime trends, for the previous month, and for the arrondissement. (Central commissaire for the arrondissement, 147)

67Here is the statistical table at the basis of these meetings. This time, we compare the results obtained from one month, within an interval of one year, again for a number of parameters of activity (line) and for the different units of the commissariat (column): the Police Brigades of the Neighborhood (BPQ) and crews from the Patrol Service.

Figure 7 : Tableau d'activité utilisé dans un commissariat central d'arrondissement en octobre 2004.

Figure 7 : Tableau d'activité utilisé dans un commissariat central d'arrondissement en octobre 2004.

68How does the commissaire use this table? The variable that interests him the most is the one that is placed first: arrests and those that interest him the least are the tickets (green signifies parking, red signifies red lights and driving, and yellow signifies speeding). For the latter, his goal is not for his officers to do as much as possible, but just enough to prevent the statistics from falling. For the most important variables, he compares the statistics from each brigade for the same month in different years, and can compare brigades with each other. He also gives them objectives-often quantified-and asks them to create the means which would help them to achieve those objectives. When the commissaire called this meeting for the first time, the men said the situation seemed a bit like a winners list, like when companies highlight the employees of the month, but they have nonetheless accepted it. Aside from that, the commissaire noted, such a table allows you not only to sanction, but also to praise those who work well and to give indicators, benchmarks, by which everyone can guide their actions. These “in-house” sessions strongly resemble prefectural meetings, if less polished-with a clearer comparative aspect. In addition, at the end of the meeting, the commissaire insists that the recommendations be transmitted to the lower ranks.

My conception of this meeting is that they echo to their supervised core employees what has been said during the meetings and instructions that I give. It is a chain of command. [...] so at the level of the basic patrolman, I hope that the instructions and remarks given at the brigade meeting were passed on to the bureaucrats there. But there is also a mechanism that is special to the PP-the roll-call assembly. You are aware that, at each shift of every unit, and so at 9:00AM when the day brigade starts its shift, the employees are gathered in the roll-call room and the chief of the brigade establishes the crews, assigns the missions to the employees in an informal manner, but all the same, it is a meeting, a roll-call assembly. (Central commissaire of the arrondissement, 147)

69The commissaire therefore relies on the fact that the intermediate level of the hierarchy effectively transmits its defined objectives, among other means, at the roll-call assembly. However, each morning, the Arrondissement Police Brigades are given, at the roll-call assembly, a report of activities (CRA), which for them amounts to a basis for work and a series of objectives to reach during the day.

So, we give them [a sheet with] objectives. [...] So they have to go through what is written there and after the rest of their activities, they’re free; in fact, the principle on the ground is the freedom to act. They have complete freedom to act, I give them a large margin for maneuver because I believe that the foundation of a community policeman’s work is contact, okay, I assign them very specific objectives such as cracking down on a certain spot, but I responsibilize them a lot, I always require initiative from them. Because, well, they are released into the wild: there is a roll-call assembly in the morning, where we assign them their missions, we distribute this type of report, which they return to us at the end of their shift, where they inform us of everything they did. Then, on the back of the sheet, they write what else they did. (Lieutenant, an officer in the arrondissement police brigade, 3)

70Here is an example of a CRA :

Figure 8: Report of the daily activities of an arrondissement brigade

Figure 8: Report of the daily activities of an arrondissement brigade

71On the one hand, this sheet indicates the daily objectives of the patrol and, on the other, it also reports on any initiatives taken. While accountability is important and initiative is at the core of the mechanism, comparison of patrols might seem to be absent.

72However, after the patrol itself, the community police officers submit their CRA reports to the office, where a supervisor (usually a person with more seniority and knowledge than others, but not yet enough to have earned the « stripes » of a brigadier) enters the data onto a computer. There is a big data entry job day to day, and so, we create monthly tables, and these monthly tables are sent to the bureau, to the BCO. (Lieutenant, officer of the police brigade for the arrondissements, 3)

73The table in which the CRAs are recorded is made of columns representing the day of the month which is sub-divided brigades’ number, as well as lines showing the main variables of activity (note that there is a second Excel page that shows other variables: higher-risk contacts-banks, jewelry stores, PTT, etc., schools and “other contacts”). Here is how this table looks:

Figure 9: Excel sheet used by the community patrol office to summarize the daily activities of the brigades. The numbers in the first line are the days of the month, and below, brigades’ numbers (which consisted of three in this arrondissement).

Figure 9: Excel sheet used by the community patrol office to summarize the daily activities of the brigades. The numbers in the first line are the days of the month, and below, brigades’ numbers (which consisted of three in this arrondissement).

74Once gathered, they are sent every month to the BCO, which synthesizes the data and, as we have seen, sends them to SPEOAD and to the assistant of the deputy director of the territorial police. The circle is completed: the statistics from the patrols are entered into the process by which they will be used to evaluate the central commissaire. For the statistics, the production chain closes in on itself, securing the mechanism as solidly as a lock between a bicycle and the bicycle stand. (Didier et al., 2009)

  • 13 We could continue even lower down on our « hierarchical » descent, if we could call it that, to the (...)

75Thus, at all levels of the territorial police hierarchy, we find mechanisms similar to Compstat, using statistical indicators, the comparing of teams, procedures to encourage the taking of initiative and finally, the distribution of responsibility13. Most of them were not specifically developed for Compstat, but the quantitative tools that already existed could quite easily be assembled to produce the same effects as Compstat. Far from reserving the taking of initiative for a narrow elite, Compstat actually distributes it broadly. Accountability, the generous girl, seems to offer herself to all. But how could it be possible for everyone to take initiative without this situation not creating conflicts of authority or contradictory orders?

76The answer can be found in the fact that each table reflects a change of scale relative to the previous one. It determines a geographical zone (the brigade measures an AIRE, a central commissaire manages all of his arrondissement, and a chief all of his sector) and a temporality (the measures of activity can be taken at a quicker pace for the lower echelons-the only constraint, for the top level, is that the timeframes used at the lower echelons can easily be aggregated to their own scale). Yet each of these spatio-temporal frameworks can capture certain types of facts, specific objects, which elude others (Didier, 2009). The transition from one scale to another is not a simple summation of a single and unique fact for zones that are increasingly extended (robbery for an AIRE, an arrondissement, or all of Paris). The transition from one scale to the other creates types of facts specific to each one (and inversely, specific scales which are the only ones that grant access to certain facts). If we take the example of robberies, the Italian-style thieves, who move rapidly over extended areas, can only be handled at the level of the sector; whereas the drunken and aggressive night-owls can be handled by arrondissement brigades. Those are different types of robberies which can only be identified at a certain scale of security activity. Since they are part of the same legal category, all these facts can be added for the central commissaire or even for the head of sector, but the initiatives of some do not infringe upon those of others because they relate to crimes that are sociologically different. The police has its own taxonomy for labelling the different scales. For example, it distinguishes the commissaires as a being a policy and directional body deriving from the officers, who, for their part, form a body related to supervision and application. Thus, each can be given responsibility for crimes on his scale or spectrum.

77This responsibilization at every echelon explains why we so often hear this claim that statistics now assume a preponderant position within the state. As such, we cannot believe such an argument: the tables that we have presented have existed for the most part at least since the 1980s. Since that time, the quantity of statistics related to the state has not varied significantly. But in the past, most of the echelons that we have presented here were transmission belts for the sstatistics that were aggregated and used only at the top of the bureaucracy, local or national (Ferret, Ocqueteau, 1998). For the bottom of the police hierarchy, the administrative tasks that consisted of producing tables were simply seen as “time-wasting” and boring. But today the radical difference is that these statistics are also used to evaluate those people who technically produced them, at every level of the hierarchy. These people now feel the weight and the power of statistics a lot more, since they are used to evaluate them, whereas that was not the case before. Mechanisms such as Compstat give importance-and thus power-to statistics because they give officers more margins for initiative and at the same time, subject them to additional evaluation sessions.

78We thus see why it seems difficult to speak about exploitation in the sense that the impetus for initiative would not be equitably distributed. On the contrary, it seems that the leeway is arranged for the actors at all levels of the hierarchy, or at least the aim or objective of such a distribution. Each echelon told us that it offers to the one just below it the conditions for taking more important initiatives. Of course, in practice, it is possible that conflicts might appear. Still, the idea that each echelon can, at its level, act on its own, has been spread by Compstat and has thus found a new legitimacy.

79But if a sense of initiative is valued by the directors who apparently expect it of others, nothing guarantees that the subordinates would think it was to their advantage to exercise their freedom. It is very possible, on the contrary, that the latter prefer to overindulge themselves and leave to others the burden of producing statistics. To encourage officers to invest in the process, two methods, quite typical in private enterprises, were added to the implementation of Compstat: the determination of the statistical objectives and the generalization of bonuses.

VI. Motivation, objectives and bonuses

80At the same time that Compstat was implemented at the Prefecture of Police in Paris, the use of statistical targets for bureaucrats also spread as an increase in the payment of salaries as bonuses. Compstat is not intrinsically related to these tools, but they were found to be mutually compatible and re-enforcing. The problem for the observer is that objectives and bonuses are not completely explicit. There is something outrageous about asking a policeman to proceed with X number of police custodies during a certain period because one suspects that he can attain a certain result through an abuse of power; and there is, as is often the case in France, a discomfort when speaking about money. The result is that our knowledge of both of them remains fragmentary. Just the same, we would like to present them here because they allow for a better understanding of the functioning of Compstat.

First question: are there statistical objectives? Commissaire D explains that effectively, there are “objectives” associated with Compstat which allow for the “strengthening of motivation [of commissaires], to determine a kind of management by objectives” (Commissaire D, 63). These objectives are discussed and established with the commissaires during the pre-evaluation session.

Q: How does one construct an objective that is given to a commissaire?

A: Well, we have to place a boost on such an offence, which is problematic. Maybe we need to shake-up the investigations service a bit because we realize that police custodies are going down and that the arrests are quite high, so that means the procedures are a bit behind. We are looking for something that will increase police custodies, to recreate the connections to justice so that we have more deferrals, pay attention to patrols because here the stats are not good.

Q: Does the prefect create a hierarchy of priorities?

A: No, not really, the commissaire knows, coming out, [of the pre-evaluation sessions] what will be decided. (Commissaire D., 80).

81But these types of instructions, established by the commissaire and his directors during the pre-evaluation sessions, which do not seem to be translated into numbers, are not seen by everyone as being « objectives » as such. When we ask a sector chief about the formulation of objectives, here is his response:

Q: Is it during these pre-evaluation sessions that the objectives are redefined?

A: No, the goals, they are defined by the director, in a general manner, but sometimes brought up at the plenary meeting of police commissaires of the DPUP, and if not, the place where we define them, were we recode them, is the meeting of sector chiefs, which meets every Tuesday morning, with Mr. Q presiding, assisted by Mister D: so these meetings are an opportunity to remind the sector chiefs which orientations are to be implemented. (Interview with the sector chief, 97).

82For this sector chief, the objectives are not simple instructions like those revealed to us by the first commissaire. They are determined annually by the director of the PUP, and are then relayed by many more intermediaries aside from Compstat-the plenary meeting of commissaires, which is annual, or the meeting of sector chiefs, which is weekly. These director objectives are clearly quantitative, as explained by a lieutenant, head of a BCO:

Well, every year we have an objective. For example, it was 2 or 3% less of declared incidents and 8% less street crime. This was fixed by the director. And I think that we stick to that. We are good because we have to be [he searches through his tables]. We are at less than 6% now, so we’re good. And in terms of street crime, we are also better. But the problem is that they will always want us to reduce things, best not to overdo it. Because what’s not good is to be too good. If you’re too good, at some point it will be too difficult the next year. (Lieutenant, head of a BCO, 138)

  • 14 D. Dufresne (2008) confirms this point when, in Médiapart, he affirms that he received, by anonymou (...)

83Thus, police activity in Paris, as in the periphery (Ocqueteau, Pichon, 2008), functions with statistical objectives which allow both for « motivating » the commissaires and allowing the hierarchy to orient their actions14. But these objectives seem difficult to use because they can cause counter-productive effects (Gwyn, Hood, 2006). As noted by the chief lieutenant of the BCO: what’s not good is to be too good - in other words, they can be diverted and end up creating behaviors that are the exact opposite of those desired. Without a doubt, it’s another reason they’re not publicly exposed.

84Now we bring up the issue of bonuses. Their advantage is that they render very palpable, if we can say that, the value of initiatives-that is to say, they effectively encourage people to get involved in the mechanism. But the question arises regarding how they are divided among the different levels of the hierarchy. We know that for the officers, some part of the bonuses are “bonuses for collective results” given to the whole of a service. Another part are individual bonuses for meritorious personnel(Ocqueteau, 2006). But it is extremely difficult to know the rules at the level of the commissaires, where such distribution is notoriously opaque. The distributions do not seem to be made only as a result of Compstat output, even if in interviews, the parties clearly indicate that as the year progresses, Compstat’s output did have a certain and gradual influence over the decision.

85However, a recent forum reveals the existence of hierarchical tensions on this point, which very closely resemble what we have referred to above as conflicts of scale (Achispon, 2009). The author, Secretary General of the SNOP, the majority labor union of the officers, explains that a priori he has nothing against management by numbers. But he notes that we should never forget that the “good numbers” which commissaires adopt to trigger significant bonuses, are “the results of the activities of all the services”. Understand that we pass from one scale to another by aggregating the data from the lower echelon. But instead of passing bonuses to their subordinates in proportion to their efforts, “the management which they practice is too often limited to pressuring the services where only the result counts”. The commissaires, whose status here is assumed to be very different from those of officers, are thus accused of not compensating, in financial terms, the efforts of their subordinates, which would allow them to shine at their own level of the hierarchy. In other words, the officers don’t mind that we passed from one scale to another by aggregation, but then they required that the financial bonuses be, in their turn, disaggregated. Everyone can be autonomous, but the autonomy has to produce the same advantages and disadvantages for all; it is unfair that some get the benefits and others the penalties. We take notice of the superb symmetry between “pressuring”, and “responsibilizing”. That said, this is a recent public topic and until now, such public conflicts have been rare.

86So, if in theory, Compstat allows for stimulating the spread of initiatives, it was accompanied by objectives and bonuses that aimed at motivating people to take these initiatives. However, neither one nor the other are talismans that guarantee a positive effect, because the objectives can be circumvented and the bonuses can create conflicts connected to their distribution. However, the police hierarchy seems to have considered them more useful than harmful.


87The Prefecture of Police in Paris imported Compstat, a management device from New York that was borrowed from private sector tools. This device is based on something resembling a statistical ritual and is used to evaluate the initiatives of the arrondissement central commissaires who are responsible for public security. They gather quantitative data on their territory: the incidents that occur there and the means at their disposal. This mechanism allows them to take initiatives, that is to say, they invent actions to fight against crime. They then present their actions and observable results to the Prefect of Police of Paris, who evaluates them. This statistical procedure is thus not just a report on initiatives, it also serves to responsibilize the commissaires and incentivize them to take initiatives. These two ideas are natural: to use the terms native to our fieldwork. In particular, we will emphasize the fact that neither was used by Foucault in his work on neo-liberalism (2004a, 2004b).

88The commissaires of public security have accepted this procedure, doubtless owing to the finesse of prefect Proust, but also because it helped them to redefine their job (Ocqueteau, 2006). One difficulty they were confronted with was territorial constraint, which made it very unlikely they would solve “sexy cases”- brilliant results that sing the praises of the judicial police. Compstat provided another metric of excellence to the territorial commissaires: the large number of processed crimes and the competence to influence the curve of criminality in the right direction. This is a movement that only Compstat can prove, precisely because it highlights individual initiatives. As Porter had already shown, quantification re-enforced the legitimacy of an institution that was despised by its interlocutors (statistics allowed the public security police to gain legitimacy in relation to the judicial police) (Porter, 1997).

89This mechanism does not relate only to the dominant echelons of public administration. On the contrary, in the Paris police, it has spread throughout the police administration. From top to bottom, officers can apparently see their initiatives encouraged, even if this is not done without creating tension among the different levels to which people are attached, because they are responsibilized.

90It seems to us that this penetration by neo-liberal management methods and objectives quite radically transforms this sovereign activity of the state. For 20 years or so, we were accustomed, thanks mostly to the excellent efforts of Monjardet (1996, 2008), and inspired by the sociology of work, to analyze the police based on a conceptual coupling: made up of, on the one hand, the regulatory control exercised by the top of the hierarchy, and, on the other hand, at the base of the hierarchy, the search for the means to escape this control by creating margins of autonomy (see Purenne and Aust 2010 for a new illustration of this idea). But what we have seen here is a perfect opposite of this model. On the contrary, what is re-enforced by the top of the hierarchy is precisely the initiative taken by police officers, all of them, which is nothing other than a sort of reformatted autonomy; initiatives that are not limited by rules but are encouraged by accountability. An accountability which is also very different from the traditional administrative responsibility of public police officers. It is as if the top leadership of the local police had heard the recommendations developed by the sociology of work and had integrated into the core of their thought processes the concept of agent autonomy, but re-formatted according to the criterion of initiative – and, as a consequence, had replaced submission to rules with the idea of a generalized responsibility. Under the influence of capitalist management, rules and autonomy are in the process of being replaced within the public security police by initiative and responsibilization. We still have to evaluate all the aspects of this very profound upheaval, including within our sociological models.

Haut de page


Achispon D., 2009, La police nationale entre culture du "chiffre" et faiblesse du management , Le Monde, 22 octobre.

Aubusson de Cavarlay B., 1996, Les statistiques de police : méthodes de production et condition d'interprétation, Mathématiques, informatique et sciences humaines, vol. 34, n° 134, 39-61.

Cauchie J-F., Chantraine G., 2005, De l'usage du risque dans le gouvernement du crime, Champ Pénal, vol. II,

Desrosières, A., 1993, La politique des grands nombres, Paris, La Découverte.

Desrosières, A., 2008, Gouverner par les nombres, Paris, Presses de l'Ecole des Mines.

Didier E., S. Névanen, Ph. Robert, R. Zauberman, 2009, La solidité des institutions. Les statistiques de victimation de l'INSEE, Genèses, vol. 74, n°1, 128-144.

Didier E., 2009, En quoi consiste l’Amérique. Les statistiques, le New Deal et la démocratie, Paris, La Découverte.

Didier E., 2011, « L’Etat néo-libéral ment-il ? "Chanstique" et statistiques de police », Terrain, à paraître.

Dubois J-P., 1996, La responsabilité administrative, Paris, La Découverte.

Dufresne D., 2008, Médiapart révèle le « tableau de bord stratégique » de la préfecture de police de Paris, Médiapart, 28 mai.

Farrell, W. J., 1994, Reengineering and the New Crime Strategies. New York City, NYPD.

Ferret, J. et Ocqueteau F. (Eds.), 1998, Evaluer la police de proximité. Problèmes de concept et méthodes Paris, La documentation française/IHESI.

Foucault, M., 2004a, Sécurité, territoire, population. Cours au Collège de France. 1977-1978, Paris, Gallimard/Seuil.

Foucault, M., 2004b, Naissance de la biopolitique. Cours au Collège de France. 1978-1979, Paris, Gallimard/Seuil.

Garland D., 1997, Governmentality and the problem of crime: Foucault, criminology, sociology, Theoretical Criminology, vol. 1, n° 2, 173-214.

Gibson J., 1977, The Theory of Affordances, in Shaw R., Bransford J. (Eds); Perceiving, Acting, and Knowing: Toward an Ecological Psychology, Hillsdale, Lawrence Erlbaum, 67-82.

Gwyn B., Hood Ch., 2006, What's measured and what matters: targets and gaming in the English Public Health care system, Public Administration, vol. 84, n° 3, 517-538.

Hache, E., 2007, La responsabilité, une technique de gouvernementalité néo-libérale ?, Raisons politiques, 28, 4, 49-65.

Hammer M., Champy J., 1993, Le Reengineering. Paris, Dunod.

Laurens S., 2007, « "Pourquoi" et "comment" poser les questions qui fâchent ? Réflexions sur les dilemmes récurrents que posent les entretiens avec des "imposants" », Genèses, n° 69, 112-127.

Monjardet D., 1996, Ce que fait la police, Paris, La Découverte.

Monjardet D., 2008, Notes inédites sur les choses policières. 1999-2006. Paris, La Découverte.

Moreau de Bellaing C., 2009, Violences illégitimes et publicité de l'action policière, Politix, vol. 22, n° 3, 119-141

Mouhanna C., 2002, Une proximité en devenir, Paris, IHESI.

Ocqueteau F., 2004, Polices entre État et marché, Paris, Presses de Sciences Po.

Ocqueteau F., 2006, Mais qui donc dirige la police ? Sociologie des commissaires, Paris, Armand Colin.

Ocqueteau F., 2009, Ce que la gestion de la sécurité publique fait aux commissaires de police, in Demazière D., Gadéa Ch., Sociologie des groupes professionnels : acquis récents et nouveaux défis, Paris, La Découverte, 63-72.

Ocqueteau F., Pichon P., 2008, La sécurité publique à l’épreuve de le LOLF, l’exemple de la DDSP de Seine et Marne et de la CSP de Coulommiers, 2006-2007, Paris, CERSA-CNRS.

Osborne D. and Gaebler T., 1992, Reinventing Government. How the Entrepreneurial Spirit is Transforming the Public Sector, Reading MA, Addison-Wesley.

Pezet A., Ponem S., 2008, Des indicateurs pour les ministres au risque de l'illusion du contrôle, La vie des idées, février.

Porter T., 1997, Trust in Numbers. The Pursuit of Objectivity in Science and Public Life, Princeton, Princeton University Press.

Purenne A., Aust J., 2010, Piloter la police par les indicateurs ? Effets et limites des instruments de mesure des performances, Déviance et société, vol. 34, n° 1, 7-28.

Robert P. et al., 1994, Les comptes du crime. Les délinquances en France et leur mesure, Paris, L'Harmattan.

Robert P., Zauberman R. et al., 2008, L’évolution de la délinquance d’après les enquêtes de victimation. France, 1984-2005, Déviance et société, vol. 32, n° 4, 435-471.

Zauberman R., 1998, La répression des infractions routières : le gendarme comme juge, Sociologie du travail, n° 1, 43-64.

Haut de page


1 Jean-Paul Proust was the Prefect of Police of Paris from 2001 to December 2004 and Pierre Mutz from December 2004 to 2007.

2 The deputy director of the territorial police explored the idea of audits conducted during visits to arrondissement commissariats by senior hierarchy, but this method was not completely satisfactory, especially because it was too slow. Compstat, which produced a faster turnover of evaluated commissariats, thus seemed perfect for achieving the goal.

3 The arrondissements themselves are divided into AIREs (Areas of Initiative, of Responsibility and of Exchange) which define the space to which patrols are assigned. These spaces proved to be very complex to define because they are subject to multiple constraints. They were first drawn up at the beginning of the 2000s and completely corrected by 2005. They had to be few enough in number for agents to be retained (which is difficult because the vast majority of officials want to return to their native provinces as soon as they can). Basically, this number was from 6 to 8 AIREs by arrondissement, knowing that in some, the same patrol was dedicated to several AIREs, but at different hours. Then it was necessary that these AIREs have a social homogeneity-that they correspond to a residential area, businesses zone, etc. This means that some parts of Paris were not part of any AIRE: railways starting from the railway stations for example, or the Northern part of the 17th arrondissement, or unique avenues, such as the Champs-Élysées, are by themselves an AIRE. Finally, it was necessary that the map shape of the AIRE be pleasant to the eye -“sellable”- to the top of the hierarchy, which explains their tendency to be produced in a very compact form.

4 I gathered up all of my data in a working paper entitled Statistics for a new management of the police? completed in 2010 and available upon request to the author ( The citations reference gives the code name assigned to the person cited and the page of the document where the citation is found.

5 The police initiative as conceived here is very close to that which the Americans call « police discretion »; but this latter term remains a pejorative term this side of the Atlantic.

6 Symptomatically, the public that had initially been accepted in New York was replaced in France by....the order for an opinion poll among the Parisian public regarding its feelings towards the police! This is an amusing example of polling which serves to...distance the public from decision-makers, rather than bring it closer (if we accept as a reference the proximity accepted in New York). During the initial evaluation sessions, the main results of this poll were discussed IFOP (Frédéric Dabi/Laurent Tison FD-LT 1-4189), The Evaluation of the Actions of Police Headquarters and the Expectations of Parisians Regarding Security. Synthesis of Results, January 7, 2004.

7 Bad luck, I wasn’t aware of current practices and am one of the first to shake his hand: Hello sir, I say, I am a researcher at CNRS etc., just after that I hear all the police officers, one after the other, say: My respects, Mr. Prefect. I understand that I have made a mistake of etiquette: we don’t greet the Prefect of Police of Paris, we offer our respect. This mistake was also noticed by the commissaire who had invited me, who said to me: OK, well, that’s it, you have now been categorized!

8 And it is difficult to describe the extent to which the sociologist, intimidated and hiding inside a business suit, with some adolescence traits, really feels like a tiny field mouse at an annual assembly of golden eagles!

9 Liaison is the internal bulletin of the headquarters of the Paris police.

10 Here it may be necessary to provide some subtext to this question. In France, one can normally lodge a complaint in any police station, no matter where the grievance occurred. For example, many people have suffered some aggression near home but, for practical reasons have lodged a complaint near their workplace. This is especially the case in tourist areas where most people work but don’t live. In this case, the commissaire records a significant volume of complaints, but he has no chance to solve them because a solution is not within his jurisdiction. In posing this question, the director of the PUP brings out a constraint that his commissaire cannot alter, and therefore supports his commissaire.

11 E. Hache notes many examples from the feminist literature that demonstrate just how responsibilization often applies to men who covertly benefit from the work that women do, especially housework.

12 The creation of a « Club of Prefects » (Le Monde, October 13, 2009), p. 12) seems indeed motivated by changes to their profession caused by the constant demands for “results” expressed by the executive.

13 We could continue even lower down on our « hierarchical » descent, if we could call it that, to the criminals themselves, because they are also now “responsibilized” in prison (Garland, 1997; Cauchie, Chantraine, 2005).

14 D. Dufresne (2008) confirms this point when, in Médiapart, he affirms that he received, by anonymous post, a file that recapitulated the quantitative objectives of all the services of the Prefecture of Police. He published some of those pages in that article.

Haut de page

Table des illustrations

Titre Figure 1: Flowchart of DPUP in 2003.
Légende Our research covers the sub-division of the territorial police, that is the part of the organization chart beginning at the top and down to the three maps of Paris at the bottom.
Fichier image/png, 251k
Titre Figure 2: Example of a summary table discussed in a preparatory evaluation meeting.
Fichier image/jpeg, 128k
Titre Figure 3: Seating plan for the evaluation meeting
Fichier image/jpeg, 136k
Titre Figure 4: Overview of a meeting (an extract from Liaison, No. 82, 2002).
Légende During the meeting that I attended, the technician projecting the images was seated at the table and not in the middle of the room9.
Fichier image/png, 316k
Titre Figure 5: The first three slides from the presentation I attended.
Légende The first one is the summary of the sector as a whole, and we immediately see (if we have color) that the arrow relating to the third variable “violent crime against people” is red.
Fichier image/jpeg, 408k
Titre Figure 6 : Tableau d’évaluation Secteur le 15 novembre 2004.
Fichier image/jpeg, 196k
Titre Figure 7 : Tableau d'activité utilisé dans un commissariat central d'arrondissement en octobre 2004.
Fichier image/png, 137k
Titre Figure 8: Report of the daily activities of an arrondissement brigade
Fichier image/png, 179k
Titre Figure 9: Excel sheet used by the community patrol office to summarize the daily activities of the brigades. The numbers in the first line are the days of the month, and below, brigades’ numbers (which consisted of three in this arrondissement).
Fichier image/jpeg, 356k
Haut de page

Pour citer cet article

Référence électronique

Emmanuel Didier, « « Compstat » in Paris: a device for prompting police initiatives and accountability », Champ pénal/Penal field [En ligne], Vol. VIII | 2011, mis en ligne le 13 avril 2015, consulté le 25 mars 2017. URL : ; DOI : 10.4000/champpenal.9028

Haut de page


Emmanuel Didier

Visiting Professor of Sociology, Permanent researcher CRNS, Deputy Director Epidapo (UMI 3663), UCLA. Email :

Articles du même auteur

Haut de page

Droits d’auteur

© Champ pénal

Haut de page
  • cnrs
  • Logo DOAJ - Directory of Open Access Journals