Between 2000 and 2010, the state of São Paulo saw a sharp decline in homicide rates, which fell from 42.2 per 100,000 population in 2000 to 13.9 per 100,000 population in 2010.1 Following this decrease, São Paulo ranked among the lowest three states in Brazil for homicide rates.

Furthermore, the reduction of homicide incidences has continued. According to data from the Secretariat for Public Security of São Paulo rates continued to fall in the State, albeit to a lesser degree, during both 2011 and 2013, having spiked only slightly in 2012.

Despite these impressive results, studies examining ‘the case of São Paulo’ have gained little attention in international forums. This lack of attention may be due to the absence of robust assessments to explain the underlying factors behind the reduction in homicides. Alternatively, it could be due to the fact that discussions on public security are considered highly sensitive among Brazilian politicians because public perceptions about security are based on a combination of factors related to quality of life and personal experience with violence, divorced from crime indices. Unfortunately, the debate on security issues is conducted in a rather erratic fashion in Brazil, typically coming to the fore only during times of great national unrest, with the implication that a mere legislative bill can remedy the problem.

The fact is that São Paulo state boasts impressive results regarding homicide reduction, which are possibly even better than figures generally lauded at international forums for cities such as Bogotá, Medellín, Diadema and Rio de Janeiro. Clearly, this does not mean the problem is solved, but the homicide rate reduction in São Paulo should be cause for celebration, especially when set against the rates recorded in other regions of South and Central America.2

Unfortunately, there is a dearth of data and studies analyzing the possible factors and policies which have contributed to this reduction. Therefore, the objective of this study was to outline some of the actions undertaken in the State of São Paulo over the last few decades and discuss their impact on homicide reduction. It also provides data on challenges that remain.

This article is divided into three sections. The first part presents the series of actions undertaken by government which are consistently cited as having contributed to the reduction in homicides. The second part explores the question of homicides due to intentional violence, with special emphasis on the involvement of police officers as perpetrators. The third and final part provides a brief overview of progress and the challenges to further reduction of the homicide rate in São Paulo. This section also provides recommendations for future action.

Reduction of Homicides in São Paulo state

The 1990s was marked by a surge in violence in the state of São Paulo, especially with regard to homicide rates which peaked in 1990 (44.1 per 100,000 population). The decade was also marked by shocking events that led to a series of reviews of institutional practices within the police forces, particularly the military police.

In 1997, an incident known as the ‘Naval Shantytown’ involving footage aired on the ‘Jornal Nacional’ news program showing a group of military police extorting, beating, torturing and humiliating residents of Diadema (located in the metropolitan region of São Paulo), sparked unrest and revolt in the population. This resulted in a host of radical changes in the Military Police of São Paulo State. The first such change was replacement of the general command of the police in 1997.

Structural changes in the military police

With a new command in place, measures were introduced ushering in a new strategy for the battalions. This comprised of the formation of territorial units, which were each allocated a share of Military Police resources. This new strategy was grounded in the triad of Community Police, Tactical Force and Operations Coordinatorship.

The Community Police was envisaged as introducing a new working paradigm founded on the participation of the community in policing. Citizen involvement was to range from helping to define the priorities of a given territory, aiding in the assessment of the local reality on the ground, to devising joint projects with police and community collaboration (Bittner, E 2003). This spelled an important watershed. This new notion of involvement differed greatly from previous practices, which had been limited to the provision of donations and material support for the police to carry out their work. Another noteworthy point was the provision for participation of civil society through an advisory committee made up of actors such as the Center for Studies on Violence of the University of São Paulo (USP-NEV), the Justice and Peace Committee, and the Coordinatorship of the Community Security Councils.

Implementation of community police bases commenced in 1997 as a pilot project in 41 police companies of São Paulo city. This action remains in operation today and boasts over 500 community police bases throughout the state, encompassing mobile, permanent and district security units.

Concerning the matter of Operational Coordinatorship within battalions, an officer was designated in each unit to be in charge of monitoring crime rates and defining actions to reduce them. Police activities became based on information about incidents and focused on reducing crime indices. In order to achieve this goal, internal reforms devoting greater police resources to these tasks and promoting more rational distribution of these resources, both territorially and for policing programs,3 were undertaken throughout the next decade. In 2002 the deployment of police resources, previously dedicated exclusively to handling incidents, was overhauled with an aim to distribute police resources according to technical criteria.

Another transition during this period, whose scope extended beyond the triad outlined above, concerned police training. The curriculum was revamped with the inclusion of Human Rights as a compulsory subject and the practice of Defensive Shooting in the Preservation of Life – the Giraldi Method, was also adopted.

The Giraldi method changed the training approach for the use of firearms, focusing on conditioning of the police officer and on correct use of the weapon. This perspective takes account of the fact that real situations calling for the use of a weapon are stressful and split-second decisions have to be made. This makes hands-on training based on real-life situations indispensable, and the police officer is taught and conditioned to discharge their firearm only as a last resort (Senasp 2013).

Integrated work in public security

Another key element contributing to the improvement of the police forces was the proper delineation of Civil and Military Police jurisdictions. Given that policing in Brazil is shared among several distinct institutions, ensuring actions are compatible and coordinated is essential for effective outcomes. This allows the two institutions to work in unison, different teams sharing responsibilities for the same territory. The allows integrated planning and information exchange because the activities of the two police forces are interdependent.

Investment in a number of information systems, including Infocrim, a system allowing geo-referencing of incidents of crime in the state of São Paulo, has also allowed smarter operations which contributed to the reduction in homicide rates. There is a consensus among experts that reliable information is a crucial element of effective work in the field of security. For example, without data on the locations of criminal events, it is practically impossible to plan the distribution of police resources effectively. In addition, the dynamics and incidence of crimes are specific to a territory and effective actions combating particular crimes must be based on data-driven diagnoses, otherwise resources are scattered and become less effective.

Criminal investigation

A factor recognized among experts as essential in the reduction of homicides was the work of the DHPP – Department of Homicides and Protection of the Individual. This department of the Civil Police is in charge of investigating homicides in the city of São Paulo whose perpetrators are unknown. The DHPP underwent reform in 1996 to become a Department of excellence in the Civil Police of São Paulo state. In 2001, the Plan for Combatting Homicides was introduced, leading to a 700 per cent increase in the number of murderers put behind bars. The Plan included investment in strategies such as identification and imprisonment of repeat murderers (Manso, B P 2012). In 2005, the DHPP attained a 65 per cent clear-up rate for cases investigated (Benites, A 2012). The precinct of the DHPP dedicated to the investigation of massacres, i.e. multiple homicides, also played a critical role. This type of crime, commonplace in the early 2000s, achieved clear-up rates of 95 per cent in 2003 and a substantial decline in occurrence throughout the decade, demonstrating the priority given by investigative bodies to tackling this crime.

Role of municipal districts

Other public policies cited as contributing to the reduction in homicides in São Paulo are the prevention policies and the concerted efforts of the municipal districts on this front. Numerous criminological theories point to the order and contextual nature of the territory as elements determining violence (Cullen, F T and Agnew, R 2011). Policies impacting these aspects are primarily the remit of the municipal districts, therefore their involvement in efforts to curb crime is important.

Despite contributing to tackling crime, public policies and municipal programs are less well documented, making it hard to gauge their impact. Moreover, assessing the effect of these actions is further complicated by the existence of a variety of different actions scattered among the cities of São Paulo state and compounded by processes of discontinuity caused by successive changes in government. In order to be able to demonstrate the contribution of these actions among the factors explaining the fall in homicides in São Paulo state, more rigorous identification and documentation of them is required.

Arms control

The policy which has perhaps been most rigorously assessed regarding its contribution to the reduction in homicides in São Paulo is that of arms control.

Econometric studies, such as those conducted by Gabriel Hartung in 2009 and Daniel Cerqueira in 2010, point to the influence of the arms control policy on the reduction of homicides in São Paulo. Clearly gun control is not the only explanation, but it has had a major impact in the reduction seen in the state. Data derived from the health system reveal that at the start of the decade, 67.5 per cent of homicides in the state were committed using fire arms4 and other studies show that some of the homicides were committed by individuals known to the victim over interpersonal disputes. Therefore, the hypothesis assumed was that control of such weapons impacts the number of homicides. Firearms are certainly not the only cause of violence, but their wide availability in a setting like Brazil’s, strongly influenced by the culture of violence, makes them far more deadly. From a short and medium-term perspective, the control and reduction in availability of arms is therefore highly effective. The reduction in the volume of arms in circulation was the result of numerous actions such as civilian restrictions on carrying arms and also the arms amnesty5 allowing people to voluntarily hand over their guns to be destroyed, both provided for under the Statute of Disarmament; on top of the large volume of arms seized by the police forces, particularly the Military police.

All of the factors described above certainly contributed to the drop in homicides in the state of São Paulo. Unfortunately, few of them have been assessed to the extent that their respective roles in the decline were elucidated. Likewise, understanding of the current climate of violence is fundamental to help devise effective public policies. Fifteen years after the first structural changes in public security began to take affect, it is important to understand the current characteristics of the homicides in the state. Have they remained the same? What is the profile of the perpetrators? Is the same public afflicted? Are they still centered on interpersonal disputes? The purpose of the third part of this article is to cover these issues.

Intentional Homicides in São Paulo City

To produce this chapter, data collected from official Police Incident Reports recorded in São Paulo city for the period between January 2012 and June 2013 were analyzed. All records were obtained through the Access to Information Law.6 The Police Reports analyzed accounted for 89.6 per cent of all recorded reports for cases of premeditated homicide during the period and 88.9 per cent of the deaths due to police intervention7 for 2012. The information on deaths resulting from police intervention in 2013 was drawn from official data released by the Secretariat for Public Security of the State of São Paulo (SSP/SP) and provides only the total number. Although classified under intentional homicides, the Police Incident Reports of robbery with violence resulting in death were not included in the analysis, representing 6.5 per cent of intentional homicides in the city of São Paulo for the period of this study as shown in Table 1.

Table 1

Intentional homicides in São Paulo city (victims): premeditated homicides, murder for the purpose of robbery, and deaths in confrontations with police. Source: SSP/SP.

Nature of Occurrence 2012 2013 Total

1st sem 2nd sem 1st sem

Deaths due to police intervention 157 194 73 424

Victims of premeditated homicides 622 875 659 2156

Murder for purposes of robbery 56 45 77 178

TOTAL 835 1114 809 2758

The Police Incident Report is the first record logged when a crime is reported. It includes information such as the nature, place and time of occurrence; description of the perpetrator, victim and witnesses when known. A list and description of artefacts seized are also included along with an open field to hold a description of the incident. In some cases, the procedures ordered and performed by police authorities are also described. The Police Incident Report is registered by the police clerk at a Police Precinct based on the information narrated to him. The narration may be done by the individual who brought the incident to the Precinct – in this case generally a police officer – and by others involved in the incident who are present at the time of lodging the report. This is the initial source of information. Since this report is not the product of investigation, the information contained in it is preliminary and therefore should be analyzed cautiously.

In any event, it is the official record and provides a broad spectrum of information which helps produce diagnoses to elucidate the phenomenon of violence, in this case, intentional homicides.

The emphasis of this analysis is centered mainly on the universe of ‘known perpetrator,’ i.e. those cases in which the perpetrator is known at the time of lodging the incident. The definition of known or unknown perpetrator is also stated in the Police Incident Report, and for the purposes of this study is taken as the classification set forth in the Police Incident Report, even when the record, upon analysis, furnishes no data on the perpetrator.

Regarding premeditated homicides, in 2012 and 2013 there was a proportional increase in the number of reports recording ‘known perpetrator’ compared with studies for the 1990s (Lima, R S). This indicates a shift in the profile of homicides in São Paulo city and also shows there was a much greater drop in the number of premeditated homicides with unknown perpetrators. In absolute terms, there was also a reduction in homicides with known perpetrators.

The combination of programs, policies and actions outlined in the first section had a significant impact on the decline in homicides in São Paulo city. According to the data given in Table 2, the initiatives predominantly impacted those homicides with unknown perpetrators. Although this category still represents the greatest number, a shift in the profile of incidents is evident. The above data does not include the 424 cases of death attributed to police intervention, and therefore the percentage of deaths with known perpetrators is even greater. This non-inclusion is explained by the fact that the cases classified as death due to police intervention hinge on the premise that the State itself acknowledges that a public governmental agent caused the death. In this regard, understanding the involvement of police in the deaths in the city gains greater relevance, in as much as they are deaths caused by public security agents and may reflect the result of their performance.

Table 2

Homicides recorded in São Paulo city by perpetrator status.8 Source: SSP/SP.

1995 2012 2013

Perpetrator Unknown 91% 81% 76%

Perpetrator Known 9% 19% 24%

Absolute Total 4277 1220 557

Deaths due to police intervention

Since the first release of quarterly statistics by the SSP/SP in 1995, the quantitative information on individuals killed in violent confrontations with the civil and military police forces have been disclosed separately. Unfettered access to this information is fundamental to allow an external mechanism of control over police activity, enabling the trends and fluctuations in these modalities to be monitored. It is important to note that these data are not computed for the effects of calculating the rate of homicide in the state, calling for extra control so as to avoid bias in the statistics. In 2000, following the enactment of resolution no. 516/00 SSP-SP, data on clashes with police by each major region of command of the police were made freely available, allowing identification of the geographic region in the state where the incidents took place.

However, since the beginning of regular disclosure of this information, there have been changes in interpretation on what should be counted as deaths due to police intervention.9 The problem of such changes is that they can cause differences in the homicide rates for the state and also lead to misinterpretation of police performance. For instance, if the deaths caused by off-duty police officers are classified as premeditated murder then this elevates the homicide rate, whereas if they are counted as death due to police intervention then this raises the level of deadly force of the police.

Before embarking on a presentation and analysis of the data, it is important to put 2012 in context in terms of homicides in São Paulo city. This year saw a rise in the homicide rate largely owing to increased deaths in the capital during the second half of 2012. In June and July 2012, São Paulo state news programs10 reported that the assassination of six police officers in June had been ordered by members of the criminal faction called the First Command of the Capital (PCC). During the second semester, numerous multiple assassinations, more commonly referred to as massacres, and the deaths of police officers, mostly off-duty, were aired raising suspicions of a possible confrontation between police and criminals. In order to confirm this claims however, the Civil Police must investigate these murders. What is clear is that the number of police officers killed in 2012 rose 44 per cent compared to 2011, a year which marked the lowest number of police deaths in the decade (ISDP 2013).

As far as the data allows, the analysis is broken down into semesters so that the context outlined above can be better appreciated. As mentioned previously, the focus of this chapter will be on those cases with known perpetrators involving incidents of premeditated murder and deaths due to police intervention, more specifically, on killings by police officers.

Intentional homicides with known perpetrators

The dataset under analysis in this chapter is described in Table 3. The known perpetrator group contains 757 cases and comprises both premeditated murders (49 per cent) and deaths due to police intervention (51 per cent). It is important to note that for a substantial proportion of cases, just over 1/3 of total intentional homicides, the perpetrator is known. In most instances this allows the perpetrator to be characterized and additional information to be compiled, providing a better understanding of the causes and motives of the incident in question.

Table 3

Distribution of intentional homicides by nature and perpetrator status. Source: SSP/SP. Produced by the Sou da Paz (I’m for Peace) Institute.

Nature Perpetrator 2012 2013 Total

Premeditated homicides (occurrences) Unknown 64.2% 66.8% 65%

Known 15.4% 21.6% 17.2%

Deaths due to police intervention Known 20.4% 11.6% 17.8%

Total 1532 630 2162

Pooling together the cases with known perpetrators reveals a significant shift in the classification of the nature of occurrences. In 2012, deaths due to police intervention represented 57 per cent of the total for known perpetrator cases whereas in 2013 the proportion of these deaths fell to only 35 per cent. This clearly exemplifies a change in the pattern11 which may equally be associated with changes in police performance12 or attributable to differences in the way such deaths are classified when the police report is recorded.

In the case of the latter reason, this is not the first time the criteria for classifying the nature of the deaths have been changed. Examination of the data published by the Internal Investigators of the Military Police of São Paulo state reveals a change in the method used for classifying incidents as of September 2008, when occurrences falling under the category ‘Individuals killed by off-duty Military Police officers’13 were no longer recorded. However, based on the data compiled for the present study, more specifically on the analysis of Police Incident Reports of deaths due to police intervention in 2012, a significant volume of cases14 with these characteristics became apparent.

It is important to point out there are two different sources of information, namely, Police Incident Reports and the Internal Affairs Department and therefore discrepancies may exist between the two. In any event, interpretation of the classification of an incident should be standardized at least where generating statistics is concerned.

The study ‘Homicides in the city of São Paulo: diagnosis of occurrences recorded between January 2012 and June 2013’15 published by the Sou da Paz institute in December 2013 highlights the main differences in characteristics and of the victims of premeditated murder by perpetrator status. In the case of victims whose perpetrator was known, 27.3 per cent were women versus 7.1 per cent when the perpetrator was unknown. The two main characteristics featuring in the Police Incident Reports were altercations and disputes between couples. For the cases whose perpetrator was unknown, half of the police reports analyzed carried no information regarding the cause while 22.2 per cent of the cases exhibited characteristics indicative of assassination. The weapon used in the two types of case also differed. In those cases with known perpetrator, knives and other sharp objects were the most commonly used weapons (38.9 per cent) followed by fire arms (38.1 per cent) whereas in cases with unknown perpetrators, over 2/3 were committed using firearms (Bento, F and Rechenber, L 2013).

Analysis of the Police Incident Reports of premeditated homicide with known perpetrators enabled police-involved cases to be identified, showing a significant difference between 2012 and 2013. While police officers were identified as the perpetrators in only ten cases of homicide (4.4 per cent of total) during 2012, there were 38 cases in the first half of 2013 alone, representing 28.1 per cent of all cases with known perpetrators for the period. It is noteworthy that 79 per cent of these police officers were off duty at the time of the incident. This shift in pattern is extremely relevant and raises questions over the quality of the data produced and the possible interpretations drawn. The fact that there were many more police-involved shootings classified as premeditated murder in 2013 drives up the murder rate for the year, since the overall rate is calculated based purely on the number of occurrences of homicide. At the same time, the great celebration in the first half of 2013 over the reduction in deaths due to police intervention was somewhat misplaced, given that under the same classification conditions of the many occurrences, the true drop would actually have been much less. However, it should not be overlooked that even considering the same conditions, there was an effective 25 per cent reduction in the number of deaths due to police intervention in the first semester of 2013 compared to the same period in 2012.

An examination of the incidences of deaths due to police intervention in this case reveals a significant involvement of police officers as the perpetrators of intentional homicides in the city of São Paulo. Of the overall total of cases with known perpetrators, the police were responsible for 58.8 per cent of the homicides reported in 2012 and for 53.1 per cent in the first half of 2013. Consolidated analysis of all intentional homicides in São Paulo reveals that the police were responsible for 21 per cent in 2012, i.e. one in every five killings in the city was committed by a police officer. In the first semester of 2013, it was a lower proportion of 17.6 per cent. Despite the reduction observed in 2013, these data reveal a totally unacceptable volume of deaths. It is critical to determine the circumstances in which these shootings are taking place.

Police involvement in intentional homicides with known perpetrators

In order to understand the dynamics involved in police-involved shootings in the city of São Paulo, the characteristics of the incidents detailed in the Police Incident Reports were examined so as to identify new systematized information. Based on the dynamics described in the Police Incident Reports, incidents were grouped into different categories so as to create parameters for the analysis. The shootings committed by the police were grouped under four categories:

  • Death due to police intervention – this category encompassed cases involving law enforcement officers in the line of duty.
  • Result of police action – cases in which an off-duty police officer witnessed a crime and intervened in the act.
  • Response to attempted crime – situations in which an off-duty police officer was a victim of a crime such as robbery or attempted murder and reacted.
  • Others – this category includes cases such as altercation, response to attempted crime by officers in the line of duty but under cover, and those showing evidence of assassination.

Several conclusions can be drawn from the data given in Table 4. The first is that the police working in São Paulo city make greater use of deadly force, in other words the use of firearms, causing fatalities. In 2012, 71.1 per cent of police-involved deaths had this characteristic, taking place during police activity. As an analytical exercise, it could be assumed that the incidents recorded as deaths due to police intervention in the first half of 2013 all took place during police activity, giving a percentage of intentional homicides of 65.8 per cent for the period.16 Analyzing whether deadly force is being used correctly is beyond the scope of this article, not least because of the source used for data collection. However, assessing police conduct shows that police activity leads to one in every seven intentional homicides in the city, a figure warranting attention.

Table 4

Distribution of intentional homicides by police officers according to modality. Source: SSP/SP. Produced by the Sou da Paz (I’m for Peace) Institute. N/A: information not available. PH: Premeditated Homicide. DDPI: Death Due to Police Intervention.

2012 2013


Deaths due to police intervention 71.1% N/A

Result of police action 1.6% 6.8% 13.5% N/A

Response to attempted crime 0.6% 18.0% 19.8% N/A

Others 0.9% 0.9% 0.9% N/A

TOTAL 322 111

The second conclusion is the impact of responses to attempted crime. Deaths resulting from these responses account for 20 per cent of all deaths caused by police officers. It is often said that if a criminal discovers they are dealing with a cop they will kill them. This phenomenon however needs to be investigated in more depth. Does the assassination stem from the mere fact the person is a police officer or from the response a police officer is likely to have in a crime situation? In order to answer this question the dynamics involved in police victimization must be explored in greater detail considering different types of crime addressed by other studies. The occurrences classified as ‘result of police action,’ although lesser in number, also give cause for concern. In engaging in this conduct, the police officer puts his own life at risk as well as those of third parties. In addition, this violates all the procedures for use of force stated by corporation and does not have the institutional support that has when it is in the line duty.

The third observation to be made relates to the classification applied in different cases. It is evident that there is a discrepancy in classification of incidents resulting from police action and response to attempted crime. When deaths with these characteristics occurred in 2012, they were classified predominantly as being due to police activity, even though they should have been classified as premeditated homicide. In 2013 on the other hand, a large volume of deaths with these characteristics17 were classified as premeditated homicide. Unfortunately, there was no access to Police Incident Reports of deaths due to police intervention during the first semester of 2013 in order to check the changes in classification. However, it is possible to make this inference owing to the major public debate which developed concerning deadly action by police in the second semester of 2012. After the change in the top echelons of Public Security in the state in November 2012, a resolution was enacted decreeing that medical assistance in cases of homicide, murder for the purpose of robbery, and deaths due to police intervention must be provided only by the specialized health service and not by police officers. Furthermore, the same resolution determined that incidents of this kind must be designated ‘deaths due to police intervention’ and no longer ‘resistance followed by death’ as it was hitherto classified. This resolution appears to have led to changes in the classifications which need to be followed in the long term to assure they are maintained.


Within a single decade São Paulo state managed to both reverse the growing trend in homicides and also promote a persistent drop in rates for the state. However, the identification of which policies and programs contributed and how remains elusive. But, it is reasonable to affirm that improvements in the police institutions played a key role in this outcome. It is also clear that achieving an effective outcome hinges on a broad array of policies which can act on the many different problems identified – violence is a complex phenomenon.

Furthermore, despite improvements, the homicide rate for the São Paulo state remains far higher than the global average and is even more striking when compared to developed nations. Besides providing a review of the main policies which led to a reduction in homicide in the state overall, this article sought to provide a new perspective on intentional homicides in the city of São Paulo, which account for 31.6 per cent of intentional homicides in the state, in an effort to contribute toward more robust and effective public policymaking.

The analysis centered on intentional homicides with known perpetrators occurring in the city of São Paulo. Special emphasis was placed on the involvement of police as perpetrators in a significant proportion of these deaths. The quality of the data, particularly regarding the classification of the incidents according to their characteristics, was also considered.

The high degree of police involvement in intentional homicides indicates major challenges ahead within the police institutions of São Paulo state, despite the many advances made from 2000 to 2010. The model of what it is to be a police officer, which is intimately associated with the institutional culture of the police forces and particularly the Military Police (Neme, C 1999), needs to be revised. There is an expectation, both among the police and the public, that a police officer must intervene using the highest level of force in any crime situation, twenty-four hours a day, even when not on the job.

Police officers working in the line of duty must follow procedures and operate within a framework which allows them to act in tough situations. They must have adequate training and have the proper equipment and communication systems to enable them to call for back up, thereby increasing the safety of themselves, fellow officers and other parties. For example, under the policing programs run by the Military Police of São Paulo State (PMESP), the operations teams always work in groups of at least two police officers. This increases the chances of a successful intervention, since there is cover at all times. Furthermore, the use of deadly force should always be a last resort pursued by police.

While this is the official line of the PMESP and efforts are clearly being made on this front, the data shows that this has not yet been successful. It is unacceptable that in a democratic society, and a city as developed as São Paulo, that the police should be responsible for one out of every seven intentional homicides in the city.

Another issue that warrants attention is the conduct of off-duty police officers. When off-duty police officers take it upon themselves to act, they endanger both their own lives as well as those of others. This mode of conduct has led to a large number of deaths in São Paulo city, including a high level of police victimization.18 Unfortunately, this behavior is lauded as ‘heroism’ both inside and outside the institution and the question of whether an off-duty police officer should be armed remains a matter of debate.

The second priority identified by this analysis is the production of quality information. The state of São Paulo was a pioneer in the practice of reporting crime statistics and this effort should be acknowledged. However, this study showed that there is still much work to be done. Inconsistent interpretation and classification of incidents of crime can lead to misleading analyses of the phenomena when the data is compiled. This data bias can lead to ineffective public policymaking, compromise the quality of indicators and promote distrust.

Police Incidence Reports currently serve as a source of information for at least two different purposes. Firstly, these records are the initial source for investigating the crime in question. Secondly, the data are also used to generate statistical data. The quality of information currently collected falls well short of what is required to perform both these functions effectively. Major investment is needed to train personnel and create a better system for data collection. At the very least, basic parameters should be defined and adopted by all civil police of São Paulo state who are responsible for the creation of Police Incident Reports and the classification of different crime events standardized.

Naturally, the actions required to further reduce intentional homicides in the city and state go beyond those outlined in this article. However, the measures proposed here represent preliminary steps that can pave the way for future improvements.