Brazil has the unenviable reputation of being one of the most violent countries in the world. Between 2003 and 2007 more than 240,000 people were murdered in the country, about 27 homicides for every 100,000 people (Carneiro, 2010a).

Within this context, as Richardson and Kirsten (2005) have pointed out, the situation in Rio de Janeiro deserves particular attention. The city witnessed an explosion of violence from the beginning of the 1980s onwards which continued until very recently and, as a major global tourist destination, the host city of the football world-cup final in 2014 and the Olympic games in 2016, the issue of crime has always received more attention in relation to Rio than it has in relation to other cities in Brazil.

The case of Rio de Janeiro, however, is also interesting because of the innovative policies that are being implemented since the end of 2008 by the governor of the state of Rio de Janeiro, Sergio Cabral, in order to at least begin to address this problem, and whose results so far have shown a significant reduction in homicide rates in the city, as official figures from the Instituto de Segurança Pública have shown (

This paper aims to investigate these initiatives from within the conceptual framework of Complexity, as defined by Cowan, Pines and Metzer (1994). Particular focus here will be given to the installation of so-called ‘Unidades Policiais Pacificadores’ (UPPs, Pacifying Police Units) in some of Rio’s most violent shanty towns. It will be argued that one of the key reasons for the success of these policies is the fact that they, consciously or not, conceive of the problem of violence in Rio de Janeiro as a Complex-Adaptive System, defined by Dooley (1997) as ‘a collection of semi-autonomous agents with the freedom to interact in unpredictable ways whose interactions over time and space generate system-wide patterns.’ These systems developed through a process of self-organization, defined by Eoyang (2001) as a process where the ‘internal dynamics of a system generate system-wide patterns.’ As such, instead of trying to ‘solve’ the problem through a top-down process, the installation of UPPs has been conceived as a way of changing the patterns of self-organization in affected areas, allowing for the emergence of new processes of development, responding to the local boundary conditions of each particular area.

Drawing on a mixture of academic analysis and interviews with elite-actors as well as some of those who implement the policy on the ground, the article will, first, outline the broad context within which the policy was developed. Following on, it will subject the policies to a Complexity-inspired analysis. This application will then allow for suggestions about future steps to be taken in order to take the policy further. Finally, the broader lessons of this policy for public policy-making and avenues for future research will be assessed.

The Context: Rio De Janeiro And The Problem Of Violent Crime

As Carneiro (2010a) points out, in 1975, the murder rate in Rio de Janeiro did not reach 15 per 100,000 inhabitants. By 1995, it had reached 64.9, making it one of the most violent cities on the planet. Over the years, the impact of this increase has been profound. The city has innumerous ‘favelas’, or shanty towns, dominance over which is often disputed between heavily-armed drug gangs and/or militias whose frequent violent confrontations with each other or the police kill hundreds of innocent people every year. According to Moser et al. (2005), in 2003, Rio accounted for 19% of all loss of disability-adjusted life years in the country.

There is an extensive literature investigating this phenomenon which points to a number of causes. As Carneiro (2010a) has pointed out, some have to do with simple political incompetence. This incompetence is linked to ‘traditional’ problems such as corruption and the ‘reach’ of the state. Within this context, the performance of the military police has received particular attention. Carneiro (2010b) identifies 4 principal problems which have seriously undermined the ability of the police force to effectively fight violent crime in the city. One is internal police corruption. Second, there is a general problem of indiscipline within police ranks ‘with orders simply not being followed’ (ibid: 64). Third, poor training leads to poor operational conduct and, lastly, there simply are not enough policemen to deal with the multitude of issues within often very difficult physical terrain. Linked to these particular institutional problems are deep structural ones of both the Brazilian economy and society at large which are very evident (and visible) in Rio de Janeiro.

This last point refers mainly to the vastly unequal distribution of wealth in the country. Brazil is one of the most unequal countries in the world and Rio, one of the country’s wealthiest cities, displays this inequality graphically. The biggest shanty-town in Rio, Rocinha, is separated from a wealthy neighboring community by one road which marks ‘a 9-fold difference in employment, a whopping 17-fold difference in income, and a 13-year difference in life expectancy’ (Goldstein & Zeidan 2009: 288). As Leu (2008) has stated, there is also a clear physical division between these areas, with most shanty towns lying on hillsides, some of which directly overlook Rio’s wealthiest neighborhoods. As will be shown below, this particular feature of Rio de Janeiro has had a significant impact on the policies adopted in the city both in the past and the present.

Yet, as shown by Guimarães, such inequality has been a structural feature of Brazilian society for literally centuries and is the results of ‘formal and informal mechanisms which preserve the existing power structures in all its forms’ in particular the police, the judicial and political, as well as the educational system (Guimarães, 2008: 16).

These structural factors combined at the start of the 1980s with some specific circumstances to produce a steep increase in violence: ‘The combination of a weak state, economic crisis inherited from the military which led to harsh economic restructuring, and the expansion of the drugs trade, led to an increase in violent crime’ (Leu, 2008: 3). The combination of increasingly harsh economic conditions led to rapid urbanization, in particular the unplanned expansion of both Rio de Janeiro and São Paulo as people migrated from north to south in search of work. This led to ‘favelalization’ of city space, manifested in both the growth of existing shanty towns and the creation of new ones, a factor which has been seen as key to violence in Rio in some studies, such as Richardson and Kirsten’s (2005).

At the same time there was a massive expansion of the drugs trade in the city which was often organized by armed groups which, taking advantage of the state’s weaknesses, installed themselves in favelas and established alternative power structures for their particular areas, underpinned by its own ‘laws’ and ‘code of conduct’ (breaks of which were often summarily punished through torture or death), the provision of some basic social services and other functions traditionally regulated by the state, such as television or the internet, as Carneiro (2010b) has shown.

Analyzing Violence In Rio De Janeiro: Complicated Or Complex?

The problem of violence in Rio de Janeiro, then, is certainly ‘complicated’. In other words, many different factors have contributed to the rise of violent crime in the city over the last few decades. This is particularly true in this case because the rise in violence also took place during a period of political transformation, with the change from a military dictatorship towards a, by now, reasonably solid democratic system, followed from the second half of the 1990s by a sustained period of economic growth. Within this context, there was a long-lasting and intense debate-one vividly recounted by Le Monde Diplomatique Brasil (2009)—amongst the political elite of Rio de Janeiro about the principal objectives of police work and the right way of transforming especially the military police from an institution protecting the state towards one whose priority is the protection of the population.

However, the view of the problem of violence in Rio being ‘complicated’ had serious consequences for the development and application of policy. These policies can broadly be divided into two approaches: One can be termed ‘containment’ whilst the other can best be described as ‘confrontation’. It will be worth looking briefly at both of these strategies before the application of Complexity and explication of current policy.

‘Containment’ was basically a policy of limiting police incursions and, therefore, involvement in those areas dominated by drug traffickers. Arguing that a policy of confrontation would lead to the loss of innocent life, the principal objective of the policy was to contain the expansion of the control drug traffickers would exercise within the city to those areas already under their control. Mostly associated with the ex-governor of Rio, Leone Brizola (1983-86 and 1991-94), whose governorship has been analyzed in great detail by Sento-Sé (2002), this policy was part of an attempt to re-frame criminality as a social problem which deserved consideration under the umbrella of re-democratization and human rights. The physical and geographical structure of Rio’s shanty towns, according to this argument, significantly increased the risks to civilian lives in case of frequent incursions. As such, non-confrontation preserved the human rights of the general population, just as containment preserved the peace in significant parts of the city.

Yet, this policy had several unintended consequences. According to Carneiro (2010b), the absence of the state allowed the drug-gangs to establish their control over virtually all the big favelas of the greater Rio area which diminished the chances of the police acting effectively. In other words, containment led to a disproportionate growth of the power and influence of the groups being contained by allowing them to establish tightly-knit nets of contact and an infrastructure for the distribution of drugs and arms which greatly enhanced their efficiency whilst reducing costs. This, in conjunction with a huge increase in the consumption of illicit drugs in Rio, especially cocaine, increased the power of these groups exponentially. As shown in Carneiro’s (2010a) overview, if one adds in police corruption (in itself fuelled by rising drug-generated profits) and an inefficient prison system—which allowed imprisoned drug-gang leaders to continue organizing their respective groups—it soon became obvious that far from containing violence, the policy actually spread the problem.

The almost logical consequence was the emergence of the opposite policy option: armed confrontation with drug gangs. Developed by governor Marcello Alencar (1995-99), the policy created incentives for the military police to kill as many drug dealers as possible, linking pay rises and promotion to the number of kills a policeman could prove. During Alencar’s term in office, the number of weapons seized from favelas rose more than 20%, whilst the quantity of drugs seized shot up by over two thirds (Carneiro, 2010b). For a time, this policy seemed to have the desired effect with the number of homicides declining for the first time since the early 1980s. At the same time, the number of innocent civilians killed also rose sharply, somewhat putting into question the overall impact of these policies on violence in the city. Nor did the policy have any appreciable effect on the levels of drug consumption and therefore drug trading. Dead drug dealers were simply replaced by new ones, pointing to another key difficulty confronting policy-makers: the persistence of often abject poverty and lack of economic opportunities which allowed for ‘drug dealing’ to remain an attractive alternative for many in their attempts to earn money, as shown by Zaverucha (2001).

These two policy approaches competed for several years, depending on the outcome of elections and therefore the occupant of the governors’ mansion. In some cases, a combination of the two approaches was applied, confronting drug dealers whilst trying to ‘urbanize’ particular favelas in order to guarantee the provision of basic social services, as in the case of the program Favela Bairro, analyzed in detail by Goldstein & Zeidan (2009).

Yet, why did none of these policies achieve its ultimate aim of consistently breaking the cycle of violence? At this point the application of Complexity will be useful.

The key problem with either policy was that they conceived of the problem confronted as ‘complicated’ as opposed to ‘complex’. In a complicated system ‘it is possible to work out solutions and implement them’ (Chapman, 2002). In one policy (containment), the solution was the physical containment and concentration of drug dealers in particular areas. This would lead to the rest of the city living in relative peace. The key determinant variables, then, were the geographical location of the problem as well as the group of people being responsible for the problem (drug dealers). In the second policy, the solution was the physical elimination of drug dealers. At the same time, the policy also had a specific focus on particular areas (the favelas), essentially leaving the rest of the city untouched.

There were, hence, significant similarities between the two policy approaches. Both isolated clearly definable variables which could then be altered or, at least, controlled which would lead to the resolution of the problem identified. As such, the policy solutions developed conceived of violence in Rio de Janeiro as a classic linear system, even though the approaches arrived at different conclusions as to what this signifies in policy terms. In other words, either approach was based on 4 basic principles:

  • Order: known causes lead to known effects under all circumstances;

  • Reductionism: By observing the behavior of its parts, the behavior of a system could be understood clockwork fashion, free of surprises. The whole was the sum of its parts;

  • Predictability: Once the behavior of a system is understood, the future course of events can be predicted by application of the appropriate inputs to the model, and;

  • Determinism: Processes flow along orderly and predictable lines with clear beginnings and rational ends (adapted from Geyer and Mackintosh with Lehmann (2005: 34).

Yet, the problem of violence in Rio de Janeiro is not merely complicated, but complex. That is to say that the key characteristics of the problem are the following:

  • A number of elements or phenomena;

  • Emergence and sensitivity to initial conditions. Its development is at best partially predictable;

  • Parts of the system are reducible whilst others are not;

  • The elements of the system form coherent patterns over time, and;

  • The system is open to its environment and therefore capable of adaptation and survival (ibid).

This means that ‘the relationship between cause and effect is uncertain and there may not be agreement on the fundamental objectives [of any given policy]’ (Chapman, 2002: Foreword). Since such systems self-organize ‘policies and interventions have unpredictable consequences [whilst complex systems] also have remarkable resilience in the face of efforts to change them’ (ibid).

It is here that one can find the reasons for the failure of the policies outlined above to deal with the problem of violent crime in Rio de Janeiro. Neither conceived of the system they were dealing with as a self-organizing Complex Adaptive System. Rather, it was thought that through either containment or confrontation the system could be simplified and its development could be controlled and moved in a more desirable direction. Little to no thought was given to the reactions either policy would cause, nor to the fact that the areas where these policies were applied were themselves self-organizing Complex Adaptive Systems which interacted with—but were not subordinate to—other Complex Adaptive Systems such as the city as a whole or the state government. As such, there was no recognition of the interdependence which existed between the systems and their semi-autonomous nature. In fact, the attempts to separate those areas with a problem (the favelas) and those supposedly without a problem (the rest of the city) served as a significant barrier to progress, creating a feeling of ‘us’ and ‘them’ and allowing for, some would say encouraging, the emergence of distinct identities, as Goldstein and Zeidan (2009) have demonstrated. In the terminology of Eoyang and Yellowthunder (2005), the policies applied accentuated so-called ‘containers’ (‘the police’, ‘the state’, ‘the favela’), and emphasized differences (one part of the city has a problem, one has not; one is ‘poor’, one is not; one counts with the presence of the state and one does not) at the expense of exchanges between the affected population and the ‘state’, essentially cutting lines of communications and fostering deep mistrust between the different parts and different groups of the city.

Only with the installation of the ‘Unidades Policiais Pacificafdoras (Pacifying Police Units, UPPs) has this pattern began to change.

The Unidades Policiais Pacificadoras

Sergio Cabral assumed the governorship of Rio de Janeiro at the beginning of 2007 with violent crime one of the key issues facing his administration. One member of his government directly involved with the issue of public safety put it thus in an interview: ‘Rio had gone through decades of bitty policies. A policy was implemented only to be changed by the next government. We needed to come up with a strategy which would be resilient and which could be embedded sufficiently to make a difference over the long term’. Critically, therefore, the government, from the very start, was conscious of the need to change patterns and recognized that such change may take significant amounts of time, critical components of complex systems.

It took almost a year before this strategy was developed: the installation of permanent Pacifying Police Units in some of Rio’s most violent shanty towns, staffed by newly-recruited police officers who receive specific training in community policing. The installation of these units would be announced in advance and would subsequently be augmented the application of various social initiatives, such as education, judicial services etc in what was called ‘UPPs social’, or Social Pacifying Police Units, launched in 2010 (see for more information).

According to the member of the state government interviewed, the basic objective of the installation of the UPPs was to guarantee the right of the population to move freely and enjoy their civil rights in any given part of the city. The installation of a permanent police presence in areas previously controlled by drug traffickers or militias and the announcement of the installation prior to doing so was intended to ‘minimise the risk of violent confrontations during the initial phase of the operation and, over time, establish the monopoly of force of the state’. The aim, therefore, was to ‘de-militarize’ the affected areas. In a second phase, the key objective was to establish a relationship of trust between ‘the state’ and the population. As the commander of one of the units put it: ‘Where we are, there was no effective presence of the state for 30 years so clearly there is mistrust. We need to show practical results in order to win that trust’. For this commander, practical results included taking weapons out of the area, allowing for the provision of basic services, as well as the progressive integration of the area with the ‘formal’ city. To this end, the staffing of these units by newly-trained police officers is crucial. According to one member of the strategic police command,

We needed to change our training and outlook. We are not [in these areas] to confront, we are there to provide security…to serve the community. As such, it is useful to have policemen who are new. It allows us to change patterns and perceptions, though it will take time’.

Having established the basic parameters of the policy, one critical question was the choice of were to establish UPPs, bearing in mind the resource intensiveness of keeping a fixed 24 hour police presence in a particular area and the fact that there are hundreds of shanty towns in the city. According to one member of the state government involved in the development of the policy,

We had to make strategic choices. The first area to be pacified (the shanty town Santa Marta) was a relatively small community but was a key area for the distribution of drugs and, crucially, weapons [to other regions]. So, it was crucial to [get hold of this problem]. [Since then] the choices have been made on a case-by-case basis and the policy has developed from there…We now have a “waiting list” of places wanting a UPP [so there has been] some cultural change but [all of this] takes time….The key was to break with the cycle of violence in some of these areas and create [an expectation] that the state will have control over areas [previously dominated by traffickers] and that [the state] will stay.

Until the end of 2011, 18 such police units have been established, covering 68 shanty towns, directly benefitting 315,000 people with more than 1 million benefitting indirectly in neighboring districts, the aim being the establishment of 40 such units until 2014 (see for the figures). During the 3 years of the policy, the police has established a presence in some of Rio’s most notorious shanty towns, including Latin America’s biggest, Rocinha, as well as Rio’s most violent complex, the Complexo do Alemão.

The impact of this policy has been felt on many levels. In the most practical terms, there has been a significant reduction in violent crime in the areas affected and in the city as a whole. For instance, according to the statistics collated by the Institute of Public Safety of Rio de Janeiro, the number of homicides in the city declined from 2,155 in 2009 to 1,422 in 2011. In areas close to UPPs, the reduction in crime is even more pronounced, with, for instance, Botafogo, the district which contains the first UPP, experiencing a 71% drop in robberies between 2009 and 2010, as the Instituto de Segurança Pública has shown (

However, critically for the long-term success of the policy, those delivering the policy have perceived a notable change in attitude towards the police: ‘We are seen in a different light here. People used to [associate the police] with violence, now they perceive us to be here to help and to stop violence’, as one unit commander put it. Studies done by the Brazilian Institute of Social Research (IBPS) underscores this impression, with 86% of people saying UPPs had made their area ‘much better’ or ‘better’ whilst 79% said that the presence of the police had eliminated the presence of armed gangs in their area. This led to 80% of respondents saying that the image of the military police had become ‘much better’ or ‘better’ (IBPS, 2010).

According to the member of the state government, this transformation in the image of the police, whilst incomplete, was what allowed for the development and expansion of the policy: ‘It is critical. There is now an expectation that, eventually, we will have a police presence in [all places] which is a huge transformation. It puts pressure on us, but it shows that we have had an impact’.

What, though, is ‘complex’ about this policy and can Complexity be used to explain this success? The first key decision was the definition of the objectives: ‘We do not intend to stop all trafficking of drugs, nor all criminality. Our aim is to break the pattern of violence which persisted in these areas and allow for a different pattern [of development]’, according to the member of the state government. It was for this reason that the establishment of UPPs has always been announced prior to their establishment, as the government member explained:

We recognise that this will mean that some drug dealers will escape. However, outside their communities they do not have the same influence and power as they did inside. We also wanted to consciously avoid armed conflict, to show that [peace] is the aim, as opposed to killing. [We have] received criticism for that, but the important thing was to change the pattern.

In fact, the government member pointed out, one of the reasons for the delay in initiating the policy was the internal debate about this approach: ‘Some just wanted more resources for the police so that more [drug dealers] could be killed…it took a while to convince [these people] that we needed a different approach.’

It was also recognized that the pattern was different for each shanty-town, as the government member continued: ‘We studied each case exhaustively before deciding on whether to go in, with how many men and what our priorities there would be. It is critical that we know each place and can respond to its particular needs’. In other words, there was recognition of the particularities of each case, be it in geographic location, size, the influence formerly exercised by alternative power structures etc.

Thirdly, there was a conscious engagement with the local population. The commander of one of the units put it this way:

[The population] know what they need much better [than we do]. We need to respond to their wishes otherwise they will not [trust us]. There needs to be constant feedback and we have consciously engaged with [representatives of the community] to show than that we are here for them.

As such, the implementation of the policy is essentially a social activity, with each local area having significant autonomy to decide its particular needs, a key plank for any successful process of self-organization, as Rihani (2002) has shown.

Critically, therefore, the basis upon which the policy was developed was different which, in turn, influenced the particular phases which followed. It was recognized that violent crime in Rio de Janeiro was a societal pattern rather than a question of particular people or clearly definable variables. The strategic decision therefore was to try and change this pattern by changing the local boundary conditions of those areas which had a critical role in sustaining this pattern. Crucially, however, there was recognition that these strategic areas confronted often very different local boundary conditions which made engagement with the local population and, hence, de-centralisation a necessity. This informed the operational policy with its focus on local engagement and relative operational autonomy for the unit commander.

As such, one can therefore identify the three essential characteristics for any process of self-organization, as outlined by Eoyang and Yellowthunder (2005). The UPPs provide the key ‘container’ for the process of self-organization. In other words, they provide the stable order without which no coherent development can take place. In addition, there is a commitment from the state government that no area of the city should be ‘off limits’ for the state and that the presence in any territory has to be permanent. The operational autonomy given to unit commanders is essential to respond to the differences which exist across time and space and ensures that they can enact measures most appropriate for their particular unit is a critical strategic decision. However, this process of de-centralisation can only work if there is engagement with the respective local community. In other words, there must be the possibility of exchange and engagement so that these critical differences can be expressed. This applies not only to the pacified community itself but also to its immediate neighborhood: ‘We know there are key differences between favelas and the rest but we need to make sure that we see them as part of one city’, according to the member of the government. ‘We cannot separate what happens in the favela from the rest.’

Crucially, therefore, the state government, this time, did not make an attempt to simplify the policy-landscape. Rather, recognizing this complexity led to crucial strategic decisions (for instance, the need for de-centralisation) which were then followed up by crucial operational decisions, such as operational autonomy for unit commanders. In other words, there was a clear and consistent thread informing policy development and implementation, as well as awareness of the imperfect results the policy will produce. The consequence has been a reasonably successful policy in that it has changed patterns of development and expectations: ‘People want us here now and they feel safer’, according to one unit commander. According to him, this expectation of security and peace has led to new patterns of behavior and a desire on the part of some members of the old drug gangs to return to the formal labour market. One spin-off effect been the ‘legalization’ of lots of other services that people need but often did not get from the state. The same commander states:

People are actually beginning to demand that their electricity be provided by the local electricity company, that their water be provided legally, that their satellite TV be provided by a satellite TV provider. They like the fact that they have a fixed date to pay their bills rather than when a trafficker knocks on the door. It allows them to plan and shows them that the state is a good thing not a threat.

Within the context of Complexity as a conceptual framework, the above points are critical. They clearly point to the necessity of both general rules and local variety. Amongst the general rules established by this policy are: the indivisibility of city space, the monopoly of the state as the guarantor of security and other basic services, the presence of the agents of the state in all parts of the city and, therefore, the guarantee of the same constitutional rights for all parts of the population. At the same time, there is recognition of local variety. Within this context, success signifies new patterns of self-organization, as opposed to ‘solving’ the problem of drug consumption and crime. The government member agrees: ‘I think we are changing expectations and that is a good thing. However, this creates its own problems and [shows] that we need to keep working and adjusting’.

None of this means that the process is completed or that there aren’t considerable problems still to be overcome. Despite the undeniable progress, Rio is still an extremely violent city with shocking levels of poverty, corruption, a thriving illicit drugs market and some areas dominated by alternative power structures, with militias made up of corrupt policemen and other elements dominating significant areas in the west of the city, as Freitas (2010) has shown. As such, the underlying causes for the explosion of violence from 30 years ago have not been eradicated. Therefore, political continuity is crucial in order to embed allow for the changes of patterns outlined above to be embedded. However, as many others have pointed out, these problems go way beyond Rio and even way beyond particular policy areas. Mistrust of the state, corruption, as well as a general disrespect for life is part of Brazilian culture and will take decades to change, as Guimarães (2008) has explained.

At the same time there are particular problems brought about by the policy. One has to do with costs. UPPs are incredibly resource intensive, concentrating a relatively large number of policemen in a relatively small area. This, in turn, could have a serious impact on the fragile change of the relationship between the police and the population: What if the expectations of the population cannot be fulfilled? What if not every shanty town which would like an UPP can get one? What about richer parts of the city? What if some of the UPPs installed do not have the same success as the first ones? All these questions are, as yet, unanswered. Again, constant vigilance is required here. For instance, there needs to be complete rigor in punishing misbehavior of police officers so as to not undermine the early success, something which historically has not always happened, as the member of the current state government readily acknowledged: ‘We need to be strict and consistent and, historically, that has not always been the case.’

The very success of the policy is already starting to have unintended consequences: Firstly, both the strategic and the operational commander interviewed for this research have alerted to the problem of police overload in the respective areas:

We are being asked to do things that we cannot do, such as sorting out problems with light and the like, for which we are not responsible. We are asked to sort out problems with rent and many other things. The worry is that people will see us as substitute for other important groups, such as the resident associations, thereby weakening the presence of [civil society].

according to the unit commander. ‘There needs to be education about how the state works and who does what’, according to the strategic commander.

A second practical issue, according to one of the unit commanders, has been the growth of the pacified shanty towns as people move in search of security: ‘We have people moving here from all over Rio as well as [from out of state].’ This has led to problems in terms of space, access to services and, as O Globo showed in its edition of 30th May 2010, increasingly, sky-rocketing property prices and rents, a key issue for a community where the vast majority of residents are poor. As such, several other initiatives will be needed to deal with the new environment created by the success of UPPs.

Implications And Conclusions

The above list of problems is in no way exhaustive and is intended merely as an illustration of the kinds of issues which have been left either unresolved or have arisen as a result of the policy itself, illustrating the ongoing nature of the process of self-organization.

Despite these problems, however, the overall assessment of the first three years of this policy is overwhelmingly positive. Without doubt, the government has changed patterns of development in some of the most challenging areas of Rio de Janeiro and has, in the process, created a different perception of at least parts of the state and different expectations by the population in relation to at least some policy areas. This represents significant progress.

In terms of the general implications of the policy approach demonstrated by the UPPs, it is striking how they take account of some of the key principles of Complexity thinking: the focus on patterns, the type of objectives being formulated, the importance of engagement and de-centralisation to make changing these patterns a social activity. However, equally striking was the fact that none of the policy-makers and practitioners interviewed for this research ever made specific reference to Complexity as a conceptual framework to inform their policy. In fact, asked about this, all interviewees displayed a certain apprehension to get drawn into what they considered to be theoretical debates: ‘I need to deal with facts on the ground. I needed to change [these facts], so that was our starting point’, according to the strategic police commander. For him, decentralizing, opening channels of communications, being aware of differences across time and space were logical steps to be taken in response to the relative failure of previous policies which did not take account of these factors. For advocates of Complexity, therefore, one of the key lessons to be drawn in their attempts to insert themselves more and more into public policy thinking is to emphasise these concepts in relation to ‘common sense policies’ rather than as part of an elaborate theoretical debate. Complexity, as presented in this article, is a conceptual framework to inform the development and analysis of practical policies rather than a fully-coherent theory.

A second key lesson to be drawn is the need to be flexible when applying Complexity concepts to different areas of public policy and to be aware of the specific circumstances within which one is acting. In the case of Rio, for example, this means being aware, on the one hand, of the deep and mutual distrust between the state and the population whilst, on the other, being conscious of the strong paternalistic culture which exists, which, in this case, translates into an expectations that it is the authorities (whoever they may be) to make and apply the rules (on this culture, see Giambiagi, 2007). As one unit commander pointed out, it is extremely difficult to reach a point where residents also feel responsible for their area: ‘This will take years….[the population] has huge expectations and many ideas but little notion that they, too, can contribute to [realizing these ideas].’ As such, local circumstances and culture have a huge impact on how and to what extent the basic principles of Complexity are applied. Whilst, for instance, it is important to de-centralise in order to take account of local boundary conditions, such a process is different in Brazil than it would be in, say, Germany where the relationship between the population and authority exists within a different context.

This leads to a third lesson: De-centralisation, flexibility and, therefore, the granting of a degree of autonomy to local actors does not imply a lessening of the importance of the state. As has been shown by the above case, changing patterns of self-organization depends on the existence of a stable framework of rules, adherence to which also requires the state. Balancing authority and autonomy, rules and freedom, rights and responsibility is a key task for any government and, as recently shown by Friedman and Mandelbaum (2011) in relation to the United States. The exact balance between these will depend on local boundary conditions. Therefore, finding this balance is often a case of trial and error, which emphasizes again the need for flexibility.

As such, one key task for Complexity practitioners now is to apply these principles to many different policy areas and political issues. Raising awareness of the fact that adhering to Complexity-principles is not a threat to the state but merely suggests a different role for it is a key task. The above case shows that some policy-makers ‘get’ Complexity inherently in some cases. Yet, bearing in mind the uncertainty of outcome, one key area for further work is the question of how advocates of Complexity can persuade policy-makers to consistently adopt models and methods inspired by Complexity. In other words, work needs to be done to legitimise the concepts of Complexity as guides to public policy. Here further engagement with the policy-making community is critical in order to show that what is being done, in this case in Rio de Janeiro, does have sound methodological and epistemological foundations which ought to be applied more broadly.