Neither Hero nor Villain: The Determinants of Confidence in the UPP Police

Neither Hero, Nor Villain: The Determinants of Confidence in the UPP Police In this article we analyze the determinants of confidence in the Police Pacification Units ( Unidades de Polícia Pacificadora – UPPs). During their existence (2008-2018), “pacification” was represented by the media as an institutional hybrid which moved between two extremes. The police were either presented as heroes, whom people wanted to guarantee the end of shootings and ensure security in areas previously marked by violence, or as villains, rejected because of the lethality of their actions and the arbitrariness of how they dealt with people. Taking these representations as ideal types, we sought to identify the determinants of confidence in the police, with the counterpoint being a domiciliary survey carried out between 2014 and 2016 with 2000 residents distributed among twenty “pacified” favelas in the city of Rio de Janeiro. The results indicated that the respondents had ambiguous sentiments towards the UPP police, mediated by the perception that shootings were part of daily life in these communities and that the “pacification” had improved security, making the police neither heroes nor villains.


INTRODUCTION
The Police Pacification Units (Unidades de Polícia Pacificadora -UPPs) were a policing model implemented at Rio de Janeiro's favelas in between 2008 and 2018 1 , as the background of the megaevents held in the city (Cavalcanti, 2015). This enterprise, called "pacification" 2 , began with the occupation of Santa Marta favela in November 2008 (Couto, 2016). Due to the reduction of shootings as a result of the police presence in the area, it was replicated in the favelas of Batan and Cidade de Deus in 2009 (Carvalho, 2018). In subsequent years, the expansion of the project accelerated, reaching 38 UPPs in 2014 (Rodrigues, 2014).
Between 2008 and 2013, a moment when the expectations of citizens inside and outside favelas seemed to be satisfied with this new form of police work (Leite, 2012), the UPP policemen were portrayed by the media as heroes 4 . They had freed favelas from the tyranny of drug dealers, guaranteeing those living in these areas the right to come and go by ending gunfights and police violence (Cano, Borges and Ribeiro, 2012).
The following years did not bring such promising results, leading to the fear of the return to the previous situation (Cunha, 2015). This change began to be perceived due the lack of definition of a program's structural base -still undecided five years after the inauguration of its first unit (Misse, 2014). While initially the UPPs were orientated by a communitarian policing philosophy (Rodrigues and Siqueira, 2012), over time they were transformed into proximity policing (Rodrigues, 2014) through the combination of repressive methods and "community action" (Carvalho, 2018:93). Until the end of the project, their essence remained uncertain, since proximity policing did not have "a more systematic conception of its practice" (Ibidem).
The lack of a clear definition of "proximity policing" meant that the "pacification" process expanded without a detailed regulation of how police should behave (Rodrigues and Siqueira, 2012) and of what UPP's policy effectiveness indicators would be (Cano, Borges and Ribeiro, 2012). This absence meant a huge power for local commanders and for the frontline police (Rodrigues, 2014), resulting in enormous diversity among the 38 UPPs implemented (Couto, 2016).
The point of inflection of the metaphor of the police officer as a hero was the disappearance of the bricklayer Amarildo in Rocinha favela on 14 July 2013 (Leite and Farias, 2018). In turn, the strength of the narrative of the UPP as a strategy involving the "continuous militarization of territory as a form of responding to the problem of public security" increased (Miagusko, 2018:163), leading to the continuity of the logic of war, notwithstanding the constant presence of the human rights discourse (Rocha, 2018:233). The police started to be portrayed in the media as villains 5 who sought to "civilize" the urban poor through the use of violence (Leite et al., 2018). A new "regime" of life under siege was thus established (Machado da Silva, 2008), leaving the residents between the "minefield" (resulting from constant police surveillance and the drug trade, "to prevent contact between residents and the enemy side"), and "crossfires" (visible in the return of gunfights in previously "pacified" areas) (Menezes, 2018:82).
The UPPs were run down 6 after federal intervention in Rio de Janeiro was decreed in 2018, reinforcing the perspective that they had only been aimed at preparing the city for the megaevents (Misse, Vargas and Couto Filho, 2020). However, since "the failure of the project cannot be a justification to abandon them as a heuristic object" (Rocha, 2018:225), in this article we seek to understand which factors contribute to a greater or a lesser confidence in the UPP police by the residents of "pacified" favelas.
This exercise does not involve reinforcing the dichotomy of the police as a hero or as a villain, since this Manicheanism is counterproductive for the understanding of the legacy of UPPs. The proposal is to understand these two categories as ideal types 7 , since they were constructed as mediatic narratives and reinforced in the academic environment. Through the dialogue with more contemporary studies of determinants of confidence in the police, we sought to identify how the residents of favelas with UPPs positioned themselves in relation to the project. Also, we sought to identify which factors might have been worked on for the "pacification" to move towards proximity with the community (Rodrigues, 2014) in the expense of the reinforcement of the "war metaphor" -which consists in a "symbolic repertoire in which sides/groups in confrontation are enemies and extermination is one of the strategies for victory" (Leite, 2012:379).

THE DETERMINANTS OF CONFIDENCE IN THE POLICE: A REVISION OF THE LITERATURE
The English language uses two words to refer to the confidence in the police: trust, which seeks to capture interaction between individual police officers and citizens; and confidence, a set of attitudes towards the police as an institution, intimately related to people's expectations of police work (Bradford and Jackson, 2010:1). Combining these two categories, we refer to the belief that the police will be effective, fair, and exhibit values aligned to the community (Ibidem), resulting in the desire to have them close as an element guaranteeing security and inhibiting crime (Awan et al., 2019).
In the international literature, there is a certain consensus that confidence in the police can be explained by three factors: a) contact with the police (Lee et al., 2019); b) expectations about the results of their work (Skogan, 2005); and c) the sociodemographic characteristics of individuals, such as age, race, gender, and education (Cao and Wu, 2017). However, there is a hierarchy among these factors -with the most important one being the individuals contact established with the organization, which can undermine the "unquestionable belief" that the institution exists to protect citizens (Lee et al., 2019).
Contact with the police can be voluntary, when individuals use the police to resolve a problem (Skogan, 2005); or involuntary, when the subject is stopped by the police (Cao and Wu, 2017). Depending on how police treat the individuals on these occasions, confidence can increase or decrease (Bradford and Jackson, 2010).
When someone is stopped and searched, one of the principal methodologies of police work, the perception of discrimination has a profound impact on police-community relations, resulting in greater hostility, lack of confidence and suspicion of authority (Awan et al., 2019:174). Stopping people violently results in the rejection of the police, reinforcing the perspective that police practices have to be perceived as fair for institutional legitimacy not to be damaged (Skogan, 2005). Not everyone is due to have direct contact with police to form an opinion of them. In the absence of contact, gossips about how the police work in the area, widely disseminated opinion polls by the media, or televised actions constitute other elements that help to form the roots of confidence (Lee et al., 2019).
These two dimensions -direct interaction with the police and "rumors" about the police actions -mean that in places where the police traditionally act in a biased way (publicly attacking individuals with certain profiles), confidence in the institution is low (Skogan, 2005). Therefore, there is a constant rejection of the police presence in areas inhabited by marginal and excluded groups, particularly young people and ethnic minorities (Bradford and Jackson, 2010:5). A past of police violence in the peripheries means that young people (more critical of police actions) and subjects belonging to minorities tend to be afraid of being attacked by the police (Lee et al., 2019).
For this reason, people who fit the "suspect" profile tend to be stopped more often and thus to distrust the police more, rejecting their presence (Skogan, 2005). This is reinforced by the fact that their reference groups (family and friends) share long narratives of police abuse (Cao and Wu, 2017). In the peripheries of the large cities, the effect of skin color surpasses age as a determinant of confidence in the police, making race one of the most consistent factors for predicting attitudes towards the local police (Awan et al., 2019:174).
In the attempt to reconstruct the image of the police on cities' peripheries, police organizations have developed community police programs (Arias and Ungar, 2009), which consist in the decentralization of decision-making by the organizational police hierarchy; the development of local responses to locally defined problems; and on the resolution of problems to respond to the concerns of crime and disorder in the community (Braga, 2015:12). These experiences have multiplied in recent decades, with the aim of improving interaction between the police and the neighborhood, improving institutional legitimacy and increasing confidence in the police (Skogan, 2005). Some analyses indicate that, despite not being very efficient for reducing crime in the long term, this model produces immediate and sustainable impacts on police legitimacy, contributing to the trust of a traditionally suspicious population (Braga, 2015).
In Brazil, studies related to the determinants of confidence in the police have similar results to those found in the United States. Brazilian literature highlights that contact with the police reduces trust in the institution, especially when it is initiated by the police itself (Silva and Beato, 2013;Zilli and Couto, 2017;Zanetic, 2017). This occurs because the way police approach people is sufficient for citizens not to trust them (Silva and Beato, 2013:149), making the greatest effects on the police's assessment (positive or negative) stem from factors related to good or bad police action. Hence, being a victim of violence or extorsion by police officers, as well as experiences of direct contact with the police, reduce confidence in the institution and its members (Zilli and Couto, 2017:681).
Similar to international studies, the perception of efficiency in resolving problems related to violence is the variable that has the utmost positive impact on confidence in the police (Silva and Beato, 2013:118). It indicates that people want the presence of police when their actions reduce crime rates. Zanetic (2017:11) found that the capacity of police to keep streets quiet, which is an indicator of police efficiency, increased in 2.5 times confidence in them. Similarly, Oliveira Júnior (2011:19) stated that confidence in the police is explained by the assessment of police performance, and that is why poor service "has a negative effect on trust, with a 38.2% decrease in the odds of confidence".
This intrinsic relationship between the perception of legitimacy and confidence in the police presents an additional challenge in cities where police action is violent, since expectations about police actions may be due to their brutality. Moule Jr., Fox and Parry (2019) emphasize that high levels of legitimacy can contribute to the reduction of the perception that the police are militarized 8 , even when they use war tactics in daily activities. For this reason, many authors predict that a high level of confidence in the police can reverberate in support for militarized practices, which are seen as methodologies for police work and, therefore, efficient mechanisms for maintaining order.
Depending on what individuals consider to be good police actions, such as the use of violence to improve security, tactics that reinforce the war metaphor (Leite, 2012) can increase confidence in the institution, which goes against what is expected from police in a democracy (Arias and Ungar, 2009). Brazil seems to be a favorable scenario for this intricate relationship, due to its high levels of police lethality and due to the fact that 57% of Brazilians support the idea that a "good criminal is a dead criminal" (FBSP, 2017). With these numbers, it is possible to state that greater confidence in the police can be a defining element of police efficiency (Silva and Beato, 2013). After all, there is a certain consensus that politics exalting police violence are commonly put into practice because they are popular, although they are almost exclusively directed to the most poor and vulnerable (Zanetic, 2017:6).
It remains to be known whether those who suffer most from police action are also those who most support this model of militarized action.
Apparently, the answer to this question is a resounding yes, since the lower the income, the greater the confidence in the police (Oliveira Júnior, 2011). For Silva and Beato (2013:148), this intricate relationship is due to the fact that individuals with a lower socio-economic status are less critical and demanding in relation to police work and, as a result, they tend to support authoritarian behavior with greater frequency. It could be due to a perverse correlation between poverty and support for militarized tactics by the police, in the expectation that their action guarantees the control of crime in peripheral areas of large cities (Zanetic, 2017;Silva and Beato, 2013). This finding goes against the international literature, in which the negative experiences that black Americans have with police can lead to lower levels of legitimacy, which, consequently, is related to the lesser willingness to support aggressive policing strategies or tactics such as militarization (Moule Jr., Fox and Parry, 2019:168).
In Brazil, although the poorest trust the police more, there are differences in the determinants of confidence among this segment of the population. This relationship is mediated by factors such as age and skin color. National studies on confidence in the police have found that young people tend to have a lower evaluation of police work; and those who are not white demonstrate lower satisfaction with the work done by police (Oliveira Júnior, 2011:20). These results reinforce the perspective that confidence in the institution varies according to sociodemographic characteristics.
There are studies that scrutinize the relationship between police and society in the municipality of Rio de Janeiro (Muniz et al., 2018). In most of these, the militarized strategies used by the Polícia Militar to patrol poor areas are emphasized as a reinforcement of the war metaphor (Leite, 2012), meaning that residents of these areas are refractory to the presence of police (Leite et al., 2018).
The novelty of this article is to examine only the population resident in favela areas, with the aim of understanding the elements that contribute to higher confidence in the UPP police. Using the history of "pacification", we can see the UPPs as a project that aimed the construction of new relationship between the police and the community, that used recruits -something unprecedented in this kind of program. This element would lead to differentiated contacts between residents and police, with the end of the police brutality evidenced in shootings with drug dealers and in the arregos 9 that enables illegal markets. (Musumeci et al., 2013;Carvalho, 2018;Cano, Borges and Ribeiro, 2012;Misse, 2014). By treating favela's residents with respect, guaranteeing people security and reducing the odds of violence in these territories, it was believed that residents' evaluation of the police work would improve and reverberate in greater trust in the UPP police.
To guarantee the expected effects in relation to the engagement of the population with the police (Braga, 2015), the policy could not have an expiration date. In the case of the UPPs, it means that they had to last longer than the megaevents (Cavalcanti, 2015). This dimension is related to the memory of the discontinuity of other projects of community policing in favelas since re-democratization (Rodrigues and Siqueira, 2012;Hollanda, 2005). Without the guarantee that with the UPPs would be different, it would be hard for residents to collaborate with the police, given the possibility of the drug traffickers settling scores once the "pacification" had ended (Menezes, 2018).
To guide data analysis, we assumed that the UPP police were represented by the media as heroes deserving the confidence of the area's residents when the feeling of security increased; when they contributed to a greater circulation of individuals within the favela itself; and when the chance of people suffering any sort of violence fell. On the other hand, the UPP police were represented as villains, deserving the scorn of the residents when they reinforced the discrimination and abuse of traditional policing in their violent treatment of people; meaning that crime continued to be part of daily life in favelas; through the victimization of residents by the police with physical and verbal aggression; or due to the continuity of firefights, which in the end was the great promise of "pacification".
Since interactions with the police and expectations related to the police work are mediated by the socio-demographic characteristics of individuals, our assumption was that young and black citizens trusted the UPP police if they had interrupted the discriminatory approach that tended to be applied to these group (Ramos and Musumeci, 2005). For this reason, these variables needed to be considered in the analytical schemes proposed to understand this project (Diagram 1).

Diagram 1 Variables considered for the analysis of determinants of confidence in the UPP Police (Rio de Janeiro, 2014-2016)
Source: Organized by the authors.
In summary, in this article we assume that confidence in the UPP police is a result of the combination of four groups of variables, which are the following: a) interaction with the police; b) expectations related to the police work; c) dynamics of the UPPs; and d) the sociodemographic characteristics ( Figure 1). These are the variables that would make residents perceive the police in a similar way to the media's representations, sometimes as villains and sometimes as heroes.

METHODOLOGY
The data analyzed in this paper resulted from 2,000 domiciliary interviews carried out with individuals over 18 years of age, equally distributed among the first 10 favelas (called old UPPs) and the 10 most recent favelas (called recent UPPs) that received Police Pacification Units 10 . We decided to construct these two groups due to the presence of the accommodation effect, when the "novelty" of the police came to be part of daily life in favelas (Oliveira, 2014). Examining the first UPPs means dealing with the effects of the UPP policy in the locations where it had become established (Fleury, 2012), due to the understanding that the greatest problems of "pacification" were concentrated in the UPPs established in latest years (Rodrigues, 2014). The researched favelas, as well as the date of the interviews, are shown in Table 1. The respondent selection process was based on random mechanisms. We used the list of all domiciles covered by UPPs taken from the National Register of Addresses for Statistical Purposes (CNEFE) 11 and selected and drew by lots of 100 domiciles in each favela -guaranteeing a random process at all stages. The refusal rate was around 4%. With 2000 questionnaires, the data gathered had a margin of error of 3.1% with a confidence interval of 95%.
Respondents were women (54%), non-white (75%), with an average age of 45 years. On average, they had seven years of education, equivalent to incomplete fundamental education, and a family income of R$1761.45 12 , as shown in Table 2.  Due to sampling procedures, it is not possible to estimate models for each favela. Hence, we aim to understand the determinants of confidence in the UPP police as a whole, to the detriment of how this relationship was constituted in the 20 surveyed UPPs.
The statistical technique chosen for data analysis was the binomial logistic regression -in which the result was not the level of confidence, but how the likelihood of confidence is increased or reduced based on a determined phenomenon (Zanetic, 2017). To unveil the determinants of confidence in the UPP police, the original variable with five categories was recoded into just two: high confidence or low confidence (value 1); high distrust, low distrust or neutral positions (value 0).
With this recoding, the purpose of this study is to understand the variance in the odds ratio of the trust in the UPP policemen according to the phenomena that, in international and national literature, explain this relationship: a) experience of subjects with the police (approach, victimization by the police, and experience of shooting), b) expectations they place on police work (in terms of the prevention of crime and the reduction of fear) and c) sociodemographic variables (which reinforced interactions and beliefs about police actions). Considering the UPPs characteristics, a group of variables about the maturation of "pacification" was added, which were: d) length of time of the UPP at the survey date, e) belief that the project would continue after the Olympics, and f) length of residence in the community, which could indicate the memory of discontinuity of community policing projects (Rodrigues and Siqueira, 2012).
To interpret the results, we decided to emphasize the Exp(B), which symbolizes the odds ratio, which is the likelihood of confidence in the UPP police. It indicates the changes that occurs in the logit through the unitary change of the explanatory variables established in the model, showing how given characteristics contributed for increasing or decreasing the odds of confidence in the UPP police. From the generated results, in addition to dialoguing with international and national literature, we will use the ideal types of the UPP policemen (heroes and villains) to assess which one the interviewees' perception were closest to.

THE DYNAMIC OF "PACIFICATION": DOES IT MAKE SENSE TO SPEAK OF CONFIDENCE IN THE UPP POLICE?
Historically, the relationship between the police and the favelas in Rio de Janeiro had been based on tension and approximation (McCann, 2013). Since the creation of the metaphor of the "divided city" (Ventura, 1994), the police raids in peripheral areas became frequent with the aim of "retaking territory" to "guarantee order" (Couto, 2016), which contributed to situations of estrangement between the two sides (policemen and citizens) (Rodrigues, 2014). The deterioration reached its peak at the end of the 1980s (Hollanda, 2005), when the entrance of weapons and cocaine into the illicit market multiplied its profitability, meaning that the interaction between police and traficantes 13 became more violent because of the search for greater profits, on both sides (Leeds, 1996).
To attenuate this tension, since then PMERJ has sought to implement community policing projects, which consisted of the permanent occupation of favela areas to prevent violent incursions by the police (Hollanda, 2005). These approximations formed the core of experiences, such as the Integrated Community Policing Center (CIPOC) and the Police Group for Special Areas (GPAE), whose purpose were to guaran-tee, through "permanent" police occupation, the reduction of conflicts resulting in deaths (McCann, 2013) and the reduction of drug dealers' control of these areas (Rodrigues and Siqueira, 2012).
The beginning of these community policing projects, implemented by police officers who had been in the PMERJ for some years, was marked by optimism, sustained by the immediate effects of the reduction of armed violence (Albernaz et al., 2007). However, they were put at stake by cases of corruption, violent actions, and excesses in the use of force, all of which were widely reported (Cardoso, 2016). A change of government led to the end of these experiences (Carneiro, 2010). The police left the favelas worse than they entered: optimism and trust in "community policing" gave way to rejection of the police (Rodrigues and Siqueira, 2012) and, to regain lost credibility, the police once again mobilized the triad "shot, beat and bomb" that defines the traditional police culture (Rodrigues, 2014). This pendular movement (Carneiro, 2010) contributed to the institutionalization of the metaphor of war as a discursive device that has been mobilized since the 1980s to justify public safety policies that names the residents of favelas as potential enemies of the city (Leite, 2012). In this discourse, favela's residents are typified as "dangerous classes" without rights to be preserved or voices to be heard (Leite and Farias, 2018:242).
To break this vicious cycle, the UPPs started with the recognition of traficantes as armed actors who possessed a monopoly of violence in favelas, establishing a type of criminal governance (Kruijt and Koonings, 2015). To guarantee the right to come and go of the residents, it was necessary to regain state sovereignty in these territories through the constant presence of newly entered police officers in the corporation (Cano, Borges and Ribeiro, 2012), which would reduce the violence resulting from the "incursions marked by truculence and the arbitrary use of force" (Rodrigues, 2012:131). It was up to the recruits to show that the work was serious and that the UPP had come to stay (Rodrigues and Siqueira, 2012:34). For this, they needed to have a pattern of behavior distinct from the old police, which would lead to greater confidence (Carneiro, 2010).
In the survey, we found that 31% of respondents said that they had (much or little) confidence in the PMERJ police (in general), while 38% stated that they had (much or little) confidence in the UPP police. On average, the confidence in the UPP police was seven percent higher than in the PMERJ police. In Figure 1, it can be seen that the differences are much more substantial in the old UPPs: 40.9% confidence in the UPP police versus 32.3% for the non-UPP police. Even in recent UPPs, the difference is not insignificant: 35.1% have confidence in the UPP police while 29.7% have confidence in the non-UPP police. From Figure 1, we derive the hypothesis that the difference in the level of confidence in the UPP police can be related to the location in which the "pacification" was experienced. Disaggregating these percentages by UPP (Figure 2), we found that in the oldest "pacifications" distrust in the police was around 30%. In the UPPs opened in 2012, distrust in the police reached a peak of 67% in Alemão and 62% in Cerro-Corá. This result reinforces the understanding that the "pacifications" were very different from each other (Couto, 2016), and that the accommodation effect could be distinct depending on the favela in question (Fleury, 2012). The differences in the levels of confidence in the UPP police could result from how commanders and the frontline policemen constructed proximity with the population (Rodrigues, 2014) 14 . The message transmitted in the official discourse to recruits was that their work should be oriented by ostensive patrolling of the whole favela (Rodrigues and Siqueira, 2012), during which there would be no use of force and whenever it was necessary to stop or search someone, this would be done in a respectful manner, without the use of weapons (Cano, Borges and Ribeiro, 2012).
This letter of intentions tried to ensure that the police would not make use of the traditional "repertoire of war", which prioritized violent conflict with the drug trafficking gangs (Teixeira, 2015:78). They would use a "social repertoire", which prioritized preventive action, through the development of activities that aimed, amongst other objectives, at the reduction of the involvement of people with crime (Ibidem). Being more "social" meant being similar to the "proximity policing" category, which came to officially define the project years after it has begun (Rodrigues and Siqueira, 2012). The expectation was that this new approach would contribute to an increase of confidence in the UPP police.
In Table 3, we present the central tendency measures of the variables that try to portray what the UPP promised in terms of contact with the police and change in the dynamics of violence, as well as the dimensions that are related to the experience of this and other community policing projects -which could impact expectations regarding "pacification". If the UPP police fulfilled the announced perspectives about police work, guaranteeing more security, confidence in its members would increase (Lee et al., 2019), transforming them into heroes. If they continued to act like the "old police", using abuses to impose order, they would be perceived as villains.  Among the researched UPPs, 78% of respondents believed that favelas were safer after the arrival of the UPP, while 91% of respondents affirmed that, with the arrival of "pacification", residents' rights were more respected. These were two promises of the project and the reason of police being portrayed as heroes. However, for 32% of respondents the chance of being a victim of violence in the community persisted, confirming the perception that the arrival of the so-called "pacification" did not meant the end of violence, but a change in patterns of crime and conflict (Menezes, 2014:675) This perception reinforced the feeling that the police were not doing anything to avoid or even investigate the occurrence of crimes within the favela (Ibidem), which damaged the image of these police officers as heroes.
Another pillar of the UPPs would be its difference in relation to the "old police", seen as violent, corrupt, and incapable of recognizing residents of favelas as citizens (Burgos et al., 2012:74). This necessarily implied changes in the police practices (such as how they stopped and searched people, and shootings) under the penalty of transforming the hero into a villain. Personal and/or domiciliary searches, experienced by 23% of respondents, give a sample of the interaction between those who live in the favela and the UPP police. There was a constant suspicion on the part of the police that the residents were collaborating with the "drug traffic" (Menezes, 2018), and searches were carried out to prevent this connection. This kind of methodology of work resulted in gossip about how the police acted, being that "the occurrence of constant police approaches and the rumor that inhabitants are being detained frequently for the simple fact that they are circulating in the favela without documents" (Menezes, 2014:671). It ended up increasing the division between the police and the community.
Since a single negative experience can seriously harm confidence in the honesty of police operations (Bradford and Jackson, 2010), any violence practiced by the UPP police could be sufficient to destroy confidence in the project as a whole to and to transform the policemen into villains. It thus calls attention for the fact that 9% of respondents were threatened, extorted, physically or verbally assaulted by the UPP police, while another 23% knew someone who had experienced the same violence. In total, 25% of respondents experienced direct and/or vicarious victimization by the UPP police -a relatively high percentage for a project whose discourse was the "pacification" of violence in the police themselves (Cano, Borges and Ribeiro, 2012) 15 . Since news spreads in favelas faster than in other parts of the city (Menezes, 2018), it is expected that both -the direct and indirect victims of the UPP police abuses -do not trust these professions and see them as villains.
For 25% of respondents, even with the UPPs, shootings continued to occur in their communities, demonstrating the inefficiency of the project. In these cases, the police are seen as villains because they are incapable of guaranteeing a ceasefire, which was the greatest expectation created by the "pacification" (Rodrigues and Siqueira, 2012).
In terms of the project's institutionality, 49% were suspicious of UPPs' ability to last beyond the Olympics -which were supposed to be the origin of their conformation (Cavalcanti, 2015). Since other experi-ences of "community policing" had failed (Carneiro, 2010), for some respondents it was evident that the UPP would follow the same path and that everything would return "to the way it was before" once the megaevents ended (Rodrigues and Siqueira, 2012). Cunha (2015) called this feeling as the "fear of the return of fear", due to the return of shooting and the life under the rules imposed by the traficantes (Machado da Silva, 2008), with reprisals against residents who spoke with the police or supported the UPPs (Menezes, 2014:677). This feeling was potentialized by the memory of respondents who had lived for many years in the favelas covered by the survey (28 years, on average) and had experienced other PMERJ community policing programs, notably GPAE (Rodrigues and Siqueira, 2012).
Finally, since the experience of "pacification" is essentially diverse depending on the moment that the project arrived in the favela (Fleury, 2012), a control variable of where the respondent lived at the time of the survey was included: whether in areas where some policy accommodation had occurred (50% of the questionnaires), or in locations where "pacification"' had occurred in the last years of the project (50% of the questionnaires), when it was already in crisis (Rodrigues, 2016).
Considering the values shown in Table 3, we can assert that there were explicit ambiguities about the effects of the "pacification". Despite the majority of respondents saying that security had increased and that there was more respect for residents' rights, a quarter of those living in areas with UPPs stated that shootings were still part of the favela's routine and that brutal aggressions and curses by the police kept going on. It is important to assess if these oppositions will remain when all the determinants of confidence in the police are triggered, which would make them neither heroes nor villains.

HEROES OR VILLAINS: WHAT ARE THE DETERMINANTS OF CONFIDENCE IN THE UPP POLICE?
"Pacification" was a project formed through practice, since unregulated "proximity policing" without regulation was transformed into a blank check that led to the excess of attributions to the police (Misse, Vargas and Couto, 2020). To act, they should rely on the commanders' direction, good will and personal intuition (Rodrigues, 2014). Some proclaimed themselves as professionals who would take care of matters that used to belong to drug dealers, understanding that their role was to replace the traficantes in regulating social life from loud sound to domestic violence conflicts (Cecchetto, Correa and Farias, 2016). Others attributed themselves the mission of civilizing the urban poor, said to be refractory to the culture of asfalto 16 (Rocha, 2018), using force as a mediation resource (Leite and Farias, 2018).
In the reality of life in favelas, residents' support for the UPP was related to the ending of armed conflicts and stray bullets (Fleury, 2012). With less "war" in these areas, the routine of these individuals would "no longer be affected by the presence and activity of traficantes" (Leite, 2012:383), which was an enormous gain in the freedom to come and go. Since the mechanisms used for this "ceasefire" varied substantially among UPPs (Rodrigues and Siqueira, 2012), some residents praised the police and other insisted "on denunciations of violations of their civil rights and violence by police agents in UPPs, who had not really broken away from the public safety policies carried out in the previous context" (Leite, 2012:383).
It is thus impossible to assert the existence of a public policy strictly speaking (Rodrigues and Siqueira, 2012), reinforcing the perception that the 'pacification' depended directly on the interactions between residents and police (Rodrigues, 2014) and the benefits that people saw in the presence of police in the favela (Cano, Borges and Ribeiro, 2012). To understand how these factors combined and explained the likelihood of confidence in the UPP police, a binomial logistic regression model was estimated. The results can be seen in Table 4.  The variables that did not reach the statistical significance are: to be searched by UPP police; to have greater respect for residents' rights with the arrival of "pacification"; whether the UPP was old or recent; length of residence in the community; gender; skin color; level of schooling; and domiciliary income. Hence, the majority of sociodemographic characteristics did not reach statistical significance, contrary to the expectations of the national and international literature. Probably this is so because the individuals living in Rio de Janeiro's peripheries are very homogenous in terms of race, income and level of schooling. Race did not prove to be relevant, which can result from the fact that 75% of respondents were black or non-white. Age -mentioned in the literature review as an important dimension in understanding the determinants of confidence in the police (Lee et al., 2019) -had statistical significance. The results indicated that in "pacified" areas, young people were extremely critical of the "pacification" model of police action, which implied additional difficulties in the construction of a better relationship between them and the police.
According to the binomial logistics regression model, the variables that reduce the likelihood of confidence in the UPP police are: presence of shooting, even with the arrival of UPPs (reduces in 53% the odds of confidence); the perception of a high probability of suffering from violence even after "pacification" (reduces the odds in 35%); and the experience of (personal or vicarious) violence perpetrated by the UPP police (reduces the odds in 26%).
The presence of aggression in the favela (despite the promise that UPPs would end this) as well as the permanence of abusive behavior by the police (in spite of the slogan of the "new police") meant that these professionals were seen as part of the problem, bringing them closer to the portrait of villains. In this case, there is the reproduction of the stereotype of the policemen as a corrupt subject, incapable of respecting the rights of the poor and useless in preventing violence. Low confidence in the police is a key point because the policing promised by "pacification" was based on mutual trust (Rodrigues and Siqueira, 2012), since "residents valorized attitudes of approximation with the police who can, based on interpersonal relations, create a climate of trust" (Burgos et al., 2012:76). If the police acted in a violent manner, the trust would be broken and the project would collapse, meaning that the "pacification" would follow the same path as other community policing projects tested in Rio de Janeiro (Carneiro, 2010).
On the other hand, the variables that increase the likelihood of confidence in the UPP police are: believing in the continuity of the project after the Olympics (increases the odds of confidence in 44%); and a greater feeling of security after the arrival of the UPP (increases the odds in 330%). If the police are capable of fulfilling one of their essential functions -the guarantee of security -those who achieve it become trustable and their presence comes to be desired in the community (Teixeira, 2015).
The results of the binomial logistic regression indicate that confidence in the UPP police can be explained by the capacity of these professionals to make "pacification" something different from traditional policing. If the police produce more security and remain there beyond the Olympics, not reproducing the previous policies' mistakes, than they deserve confidence. On the contrary, if they reproduce the vicissitudes of the PMERJ, characterized as violent and as potentiator of shootings, they do not deserve confidence. Comparing the value of the Exp(B), we found that contact was not as important for confidence in the UPP police as satisfaction of expectations. While the presence of shooting reduced the odds of confidence in the police by 53%, the feeling that there was more security increased it by 330%. Feeling safe was thus the element that most impacted the confidence in the UPP police officers.
Given these results, we can state that the UPP police cannot be seen either as heroes or villains, with the feeling of the respondents towards "pacification" being ambivalent. We understand that the UPP police could be represented as heroes worthy of the trust of local residents when they helped to increase the feeling of security, guaranteed greater circulation of individuals within the favela and decreased the likelihood of residents suffering any kind of violence. Of these three requirements, two were met for a while. The increase in the feeling of security contributed to the affection for the UPP police, while the sensation of being liable to become a victim of violence reduced confidence.
As in the case of heroes, two dimensions of villains found resonance in the data analyzed. We understand that the UPP policemen could be represented as villains who should be despised by the residents when they represented the reincarnation of the old police, with their violent approaches, increasing the problem of crime in the life in favelas, with the victimization of the residents by the police through physical and verbal aggression -which showed statistical significance, reducing the odds of confidence in UPP police officers by 26%. They were seen as villains especially when they did not contribute to the end of shootings, which was the great promise of the project -the presence of the shootings reduces the degree of confidence in the UPP policemen by 53%.
Therefore, we can conclude that the determinants of confidence in the UPP police indicate an ambiguous feeling, as the police officers were sometimes perceived as heroes, sometimes as villains. Respondents sought approximation with the project in the perspective of having greater safety and lower chances of suffering violence, but they desired distance from this experience due to exposure to police brutality, of which shooting was the maximum expression. The fear that the "pacification" would end after the Olympics contributed to this ambivalence not to dissipate over time. In 2018, two years after the Olympic Games, we witnessed the end of the UPPs with the decree of federal intervention (Miagusko, 2018), leaving as a legacy unresolved issue in the relationship between the police and the community, and the difficulty of a new approximation, given the activation of new militarization practices (Misse, Vargas and Couto Filho, 2020).

FINAL NOTES
In this article, we seek to understand the determinants of confidence in the UPP police as a way of elucidating the elements which contributed to this program to be simultaneously desired and repelled by those who experienced it daily. In this posthumous assessment of the determinants of confidence in the UPP police, we wanted to identify the points that collaborated to a positive assessment of the police action, which can serve as the basis for the next experience of community approximation, as it tends to occur in Rio de Janeiro (McCann, 2013).
To understand the challenges that the UPP intended to overcome, we reported how historically the alternation between community policing programs and the use of war tactics in favelas occurred -that has trapped PMERJ in a vicious circle since re-democratization (Carneiro, 2010). The UPP was expected to break this pattern, but instead we saw that it reproduced this path dependence (Leite et al, 2018), as announced by the residents that compared UPP to the GPAE when the first units started to be implemented (Rodrigues and Siqueira, 2012).

Ludmila Mendonça Lopes Ribeiro, Márcio Grijó Vilarouca
The data presented in this paper indicate that the UPP police cannot be considered as heroes, in the sense of returning democracy to the favelas (Kruijt and Koonings, 2015). They often acted in an abusive manner, contrary to the project's assumptions, subjecting the residents of these areas to violence and not guaranteeing a "ceasefire" from firefights (Rodrigues and Siqueira, 2012). Nor can they be classified as villains, due to their capacity to often guarantee greater security in some favelas, with this being the result most valued by our respondents (Carneiro, 2010).
In this exercise, we were interested in verifying if the factors that contributed to the likelihood of confidence in the UPP police were the same of those presented in the national and international literature. We found that in terms of sociodemographic characteristics, only age had statistical significance, going against other studies, which pointed to the prominence of race. Probably, this result was related to the fact that the majority of respondents were black, with low incomes, and a low level of education. In this group, heterogeneity was demarcated by age and for this reason only this factor influenced the results, reinforcing the finding that young people are those most critical to police work, which implies obvious challenges for the implementation of "community policing" (Braga, 2015).
In the comparison of the factors that most affected the likelihood of confidence in UPP police, we found that the effect of feeling safer after the arrival of the UPP overcame the rejection related to the continuity of shooting. With this, we found that in the "pacification" the variables related to expectations about the police work outweigh the effects of the interaction with the police, in contrast to what was hypothesized in other studies. Such a result can be a warning about the risk of militarization reinforcement, since international research indicates that good assessments of police work help to cover up the nuances of war triggered by the "guarantee of order" (Moule Jr., Fox and Parry, 2019).
In conclusion, we believe that the UPP policemen should be seen as hybrid characters between the hero and the villain, who, while guaranteeing more security, violated the basic assumption of democracy from their violent and corrupt action -which ended up contributing to the continuity of the shootings and, today, to the war metaphor (Leite, 2012).
(Recebido para publicação em 7 de agosto de 2017) (Reapresentado em 4 de junho de 2019) (Aprovado para publicação em 3 de setembro de 2019) not the "base of drug traffic". The UPPs were initially implemented in the "quietest" favelas and for this reason the challenges were lower (Fleury, 2012), which contributed to a greater effectiveness of the project and as a result greater facility in collaborating with the police.
15. Vicarious victimization was added to individual victimization because the majority of respondents (87%) said they had not suffered violence by the UPP police, but a large part of them (23%) stated that they knew someone who passed through this experience. We believe that as it happens in research on racism, some respondents did not recognize personal experience, attributing it to others, for which reason the individual victimization percentages were added in this variable.
16. In popular jargon, the city areas that are not favelas are called asfalto (asphalt). It is a counterpoint to the favelas, which are mostly located on the city's hills (morros).

Neither Hero, Nor Villain: The Determinants of Confidence in the UPP Police
In this article we analyze the determinants of confidence in the Police Pacification Units (Unidades de Polícia Pacificadora -UPPs). During their existence (2008-2018), "pacification" was represented by the media as an institutional hybrid which moved between two extremes. The police were either presented as heroes, whom people wanted to guarantee the end of shootings and ensure security in areas previously marked by violence, or as villains, rejected because of the lethality of their actions and the arbitrariness of how they dealt with people. Taking these representations as ideal types, we sought to identify the determinants of confidence in the police, with the counterpoint being a domiciliary survey carried out between 2014 and 2016 with 2000 residents distributed among twenty "pacified" favelas in the city of Rio de Janeiro. The results indicated that the respondents had ambiguous sentiments towards the UPP police, mediated by the perception that shootings were part of daily life in these communities and that the "pacification" had improved security, making the police neither heroes nor villains.