From an offender-based to an offense-based justice: Changes in sentencing patterns in the juvenile justice system in São Paulo from 1990 to 2006

Juvenile justice systems around the globe are becoming increasingly more similar to criminal justice systems. In Brazil, previous legislations focused on the individuals themselves and did not distinguish between young offenders and children in precarious conditions, but a new legislation in 1990 marked a rupture and introduced elements of criminal law. We leverage a unique data set representative of every adolescent who has been through the juvenile justice system in the state of São Paulo between 1990 and 2006 and provide a quantitative assessment of the changes in sentencing patterns in the period. Results suggest that judges increasingly prioritise violent and drug-related offenses when convicting adolescent defendants, indicating that the Brazilian juvenile justice system progressively resembles the criminal justice rationale by emphasising the ideal of proportionality between crime and punishment. We conclude with a discussion on pendular justice, suggesting that juvenile justice in Brazil is moving from a positivist-inspired to a classic-inspire justice system.


Introduction
Adolescents who engage in criminal conduct might face distinct judicial consequences. In some countries, they might be criminally prosecuted as adults; in others, they might not be legally eligible for a criminal prosecution; but in most countries these days, young offenders would go through a specialised justice system through which they would be prosecuted as underage defendants: the juvenile justice system. The degree to which juvenile justice systems worldwide are similar or distinct from classic criminal justice systems for adults is context-dependent, but recent studies have identified an average global trend since the 1990s wherein the judicial treatment of young offenders is increasingly more punitive and repressive, resembling the treatment offered by criminal justice institutions to adults who engage in offending behaviour (Trépanier, 1999;Bailleau, 2002;Muncie, 2005Muncie, , 2008Pires, 2006;Piñero, 2006;Sallée, 2017). In this study, we focus on the Brazilian case to investigate changing sentencing patterns in the 1990s and early 2000s and the growing proximity between juvenile and criminal justice systems.
Brazil consists of an interesting case to study dynamics of the juvenile justice system. Until 1990, a couple of legislations set the tone for the treatment of young offenders -the Minors' Codes from 1927 and 1979, which were centred around the so-called minors (Alvarez, 1989). An umbrella identifier, minors included both the group of young people who engaged in criminal conduct and poor children in precarious structural conditions. The main institutions in charge of handling minors were confinement units called FEBEM-SP, which were similar to prisons but for young people -i.e., for young offenders and poor children alike. By conflating economic disadvantage and penal sanction, previous legislations and all of their institutional apparatus essentially criminalised poverty (Adorno, 1993).
This situation lasted until 1990, when a new piece of legislation -the Statute of the Child and the Adolescent (ECA) -ruptured the stigmatised notion of minors. Now completely distinguishing adolescents in conflict with the law from poor children and adolescents, the confinement institutions are now exclusively available to the latter (Paula, 2011). Specialised judges are supposed to assess whether young defendants indeed engaged in offending behaviour, and, to the extent that they specifically engaged in violent crime, convict them with a detention disposition -i.e., send them off to a confinement unit for at least six months (Oliveira, 2017).
From care units devoted to young offenders and disadvantaged children alike to detention centres in charge depriving the liberty of convicted adolescents, confinement institutions went from being the symbolic representation of minors to becoming the institutional image of juvenile punishment. Likewise, whilst previous legislations focused on the underage individual who had engaged or had the potential to engage in offending behaviour, the new legislation considers the seriousness of the offense itself as the main criteria for the deprivation of liberty, suggesting responses should be somewhat proportionate depending on how violent the crime was. This shift from an emphasis on the offender's characteristics to the seriousness of the offenses, especially prominent in custody-based dispositions, indicates the legal framework orienting the Brazilian juvenile justice system is departing from principles based on the positivist school of criminal thought towards a logic usually associated with the classic school of criminal thought (Alvarez, 2014).
In this study, we assess the extent to which sentencing patterns in the Brazilian juvenile justice system followed this pendular justice mode, moving from a focus on individuals and the causes of crime to a focus on the offense and due process. Using data representative of every adolescent who had some passage through the juvenile justice system in the state of São Paulo between 1990 and 2006, we investigate the degree to which the weight given by judges to the seriousness of the offense increased over time, implying an increasing proximity of the juvenile justice system with the criminal justice system for adults. The period we are analysing is interesting precisely because it consists of the transition from the old legislation to the new, and thus we can more accurately assess changes in sentencing patterns.
The paper proceeds as follows. We first discuss the experiences of early punishment in Brazil, with focus on discourse changes from young offenders as minors to adolescents in conflict with the law. We then move on to an overview of the international debate, which has been increasingly showing how juvenile justice systems are more and more similar to criminal justice systems. Next, we describe the particularities of the Brazilian case and the recent legislative changes. After that, we present this study, discussing our hypotheses, methods, measures, and analytic strategy. We then show the results, and include some discussions about how the Brazilian juvenile justice system has changes its sentencing patterns. We finish it off with some final remarks.

From minors to adolescents in conflict with the law: early experiences of punishment
In analysing the trajectory of Brazilian children marked by early experiences with criminal justice institutions, Adorno (1993) explored the criminalisation process whereby poor children are converted into minors, a stigmatised social identity encompassing both juvenile offenders and poor children in disadvantaged conditions. Agencies and institutions dedicated to the control and processing of children and adolescents accused of criminal conduct would be particularly crucial to the production of minors. The persistent contact with legal institutions and the long periods of confinement -a situation that was fostered by legal devices under previous legislations -would be decisive in shaping the trajectory and identity of both children in precarious conditions and juvenile delinquents. 1 In Adorno's analysis, the regular functioning of criminal justice institutions would therefore produce the so called 'minors' issue' in the country, both through the production of subjectivities and identities and through the expansion and diversification of social control strategies.
Currently, it is possible to claim that the scenario described by Adorno has partially changed. Especially with the approval of the 1990 Statute of the Child and Adolescent (ECA -Law N. 8.069 of July 13, 1990) and the articulation by actors and movements that prompted the new legislation, 2 there was a growing effort to affirm children and adolescents as legal citizens (as opposed minors, who had fewer legal rights). In particular, an effort was made to explicitly distinguish offending behaviour from structural disadvantage, with the goal of restricting juvenile incarceration and fighting the criminalisation of poverty (see Alvarez, 1989). Thus, instead of minors, the figure of "adolescents in conflict with the law" emerged, a category that is exclusive to adolescents accused of criminal conduct. Deprivation of liberty through confinement and institutionalisation, now subject to the principles of brevity and exceptionality, then became just one out of six "socio-educational dispositions" to be sentenced by a specialized judge -all of which would be exclusively applied to adolescents in conflict with the law, not to those in precarious conditions. The new legislation therefore recognises the coercive aspect of detention and establishes that this disposition be only applied in cases of violent crime, as a last resort, and for the shortest possible time. The sanctioning of adolescents should, according to this logic, privilege community measures that do not entail restriction or deprivation of liberty.
Thus, from a legal point of view, ECA brought significant changes to the treatment of young offenders, such as greater specialisation of the justice system, with measures that are specifically aimed at holding adolescents accountable, as well as restrictions to incarceration. But the scope of the legal changes remains under debate. What are the effects of these changes on the way the functioning of the juvenile justice system? Has the pattern of judicial decision-making changed since the approval of the new legislation?
Recent data on juvenile sentencing in Brazil (Brasil, 2019) indicate that confinement dispositions are actually a minority of all cases when compared to community-based measures, 3 representing 18.2% of all dispositions implemented in 2017. However, if we look at the time series of juvenile incarceration (Brasil, 2009(Brasil, , 2012(Brasil, , 2019, 4 it is possible to see that the number of institutionalised adolescents rises 3 The Statute of the Child and Adolescent defines six Socio-educational Measures that can be applied to adolescents convicted of some infractions (conduct described as a crime or criminal misdemeanor) (Art. 103): I -warning; II -order to repair the damage; III -provision of services to the community; IV -assisted freedom; V -insertion in a semi-open conditions; VI -admission to an educational establishment; VII -any of the protective measures (Art. 112; Art. 101, I to VI). The so-called "open-environment measures" are those that do not imply restriction or deprivation of liberty. Measures of semi-open conditions and institutionalisation are called "closed-environment measures". 4 This survey has been carried out since 1996 by the Secretariat for Human Rights of the Presidency of the Republic (SDH/PR), but the methodology currently adopted was only established in 2009. After the approval of the National System for Socio-Educational Support (Resolution No. 119, of December 11, 2006, of CONANDA;and Law No. 12,594, of January 18, 2012), the survey started to be used as an instrument for monitoring the system. The last issue was published in 2019 and contains data for the year 2017.

3
From an offender-based to an offense-based justice: Changes… continuously from 1996 to 2015, with a slight decrease in 2016 and 2017. The country went from 4,245 confined adolescents in 1996 to 26,109 in 2017 -an accumulated growth of 515% in the period. As such, the extent to which community-based measures have been adopted as alternative strategies to institutionalisation is not clear. How can we interpret this increase in the number of confined adolescents in conflict with the law that have been institutionalised in the period?
Such growth trends in the number of citizens deprived of their own freedom while confined in criminal justice institutions is not specific to the juvenile justice system. Data on arrested and imprisoned adults in Brazil suggest similar dynamics. Between 1995 and 2016, the Brazilian prison population went from 148,800 to 726,700, a cumulative growth of 388% 5 -which is smaller than the relative growth among adolescents, but considerably larger in absolute terms. Considering the period we are investigating (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006), the growth of the prison population consisted of 345%. Growth trajectories of the number of imprisoned citizens indicate a mass incarceration trends, which has been interpreted as part of the so-called 'punitive turn' in the Global North (Garland, 2001). When it comes to youth justice systems, however, ascending trajectories in the number of adolescents confined in justice institutions alone are not enough to suggest a punitive turn. As mentioned above, the excessive use of confinement institutions is a distinctive characteristic of minors' justice systems. Depriving young people of their freedom could be evidence on both the retributive and repressive face of a criminal-based juvenile justice system and the persistence of a historical and protective youth justice focused on minors. As such, we ask: is institutionalization once again being used in response to adolescents' social vulnerability or has the Brazilian youth justice system indeed become more punitive after the new legislation took over?
The hypothesis that juvenile justice is becoming more "punitive" and getting close to the rationale of criminal justice has been discussed by juvenile justice researchers in European and North American countries. According to several of them (See Trépanier, 1999;Bailleau, 2002;Muncie, 2005Muncie, , 2008Pires, 2006;Piñero, 2006;Sallée, 2017), specially during the 1990s, a transformation process forced this specialised justice to lose its typical characteristics. For these authors, the assistance, preventive, protective, and educational model lost ground to a punitive and security logic, with the increase of repressive measures, aimed at controlling public order and defending society against youth delinquency. More recently, scholars have identified a possible reversal of these trends from the late 1990s onwards, with a decrease in detention rates and a reduction in the use of more punitive forms of intervention such as juvenile waiver and juveniles in adult prisons (Benekos & Merlo, 2008;Merlo & Benekos, 2010;Bateman, 2012;Smith, 2014;Case & Haines, 2021). The causes, consequences and significance of these changes are still under debate. Part of the literature is sceptic at assuming this movement as evidence that heavy-handed strategies are no longer being used, and either argue that key elements of the old model persist (Goshe, 2015) or show how the new model brings new types of intrusive forms of social control (Cate, 2016). Additionally, the focus on individual responsibility and on risk management, now prominent, was already part of the contradictory orientations of the punitive period.
The extent to which the Brazilian juvenile justice system followed the transformation process identified in other countries is not clear from the formal changes outlined in the new legislation alone. Likewise, the aggregate data aforementioned on the Brazilian system are incomplete and do not allow for much progress in understanding the functioning of the juvenile justice system. This study fills this gap with a quantitative assessment of sentencing patterns in the state of São Paulo between 1990 and2006. 6 São Paulo is the state with the highest absolute number (41.8% of the total) and the fifth highest rate of adolescents in custody institutions in the country (150 per 100,000 adolescents). 7 These data refer to a particularly significant period in São Paulo's childcare system's history. The period from 1990 to 2006 8 corresponds to the institutional transition process after the approval of the new legislation. In the state of São Paulo, the process of adjusting the old institutional framework to the new legislation took 16 years. Thus, analysing the possible changes to the decision-making patterns of juvenile justice in this specific period allows not only for exploring the extent to which these changes accompany transformations identified in other countries, but also for verifying the ways in which institutional practices have adjusted to the legislative changes in the Brazilian case.
In order to ascertain if the functioning of juvenile justice in the case of São Paulo has adopted a punitive logic, we analyse sentencing patterns over time, with focus on judicial decisions mandating deprivation of liberty. The database comprises 1,581 institutional case files and the analytical strategy adopted consists of logistic regression models. We investigate both if the probability of applying detention has changed over the years, and if the weight given by court rulings to the seriousness of the offense also changed. The increase in the use of detention associated with the weight given to the seriousness of the offense could indicate a punitive orientation. This means custody is not being used as a child-welfare measure, like the old tutelary system, but is also not exceptional and limited as advocated by the protective framework of ECA.
From an offender-based to an offense-based justice: Changes…

Pendulum justice: two models of justice
As previously indicated, current research suggests that, like criminal justice systems for adults, juvenile justice systems in Europe and North America have at least partially gone through a "punitive turn" 9 (see Muncie, 2008). In the case of this specialised type of justice, however, this punitive and repressive redirection would have demanded a radical change in the rationale behind judicial practices. Whilst the model traditionally identified with juvenile justice focused on the social and psychological needs of children and adolescents and was based on the possibility of rehabilitation through individualised treatment, the operating logic during the 1990s was guided by the seriousness of the crime, by the need for holding adolescents accountable for their actions, by the intention to incapacitate the offender and protect society. One of the results of this change was an internal specialisation within juvenile justice that went on to explicitly distinguish cases of abandonment and neglect from adolescents accused of criminal conduct -and juvenile justice systems worldwide starts to exclusively handle the latter case (Bailleau, 2002;Piñero, 2006).
From a discursive point of view, this substantive change was possible due to the emergence of severe criticism of the old model, both because of its inefficiency in tackling youth crime and because of how children's rights were excessively restricted, resulting in informal procedures, discretionary decisions, and lack of procedural guarantees (See Trépanier, 1999;Muncie, 2005;Feld, 1997;Bailleau, 2002). Amongst the main factors associated with these criticisms is the widespread disenchantment with the rehabilitation ideal and the subsequent disbelief that specialised agencies can prevent crime and recidivism (Trépanier, 1999, p. 318).
It is possible to say that the axis from which these transformations have been organised was the change in the focus of juvenile justice: from the offender -their individual needs, their problems, the causes of their behaviour, and the appropriate measures to deal with them -towards the offense -its seriousness, the danger it represents to society, the harm it does to the victim. Favouring the offense as the main criterion for judicial decisions meets two demands which are, at most, contradictory: on the one hand, to expand the repression and severity of punishment against adolescents who commit crimes and, on the other, to ensure respect for their individual rights by adopting more "objective" criteria in decisions and by introducing procedural guarantees. As Garland (1999Garland ( , 2008 demonstrates in his discussion of the recent transformation process in penal policies and the functioning of criminal justice, the "crisis" of the rehabilitation ideal and of what he called penal welfarism involved different demands and resulted in ambiguous and often contradictory policies. The changes observed in juvenile justice systems during the 1990s could be analysed based on two distinctive traditions of criminal theory: the "classic" school and the "positivist" school of criminal thought. These changes speak to the replacement of the notion of the offender as someone with biological inclinations or as a victim of social environments (positivist school of thought) by the notion of individual responsibility for the crime committed (classic school of thought); to claims that response to crime should no longer be measured by the need to treat the offender (positivist school of thought), but by the severity of the offense (classic school of thought); to a justice system which, rather than offering preventive treatment to the causes of delinquency (positivist school of thought), deals exclusively with adolescents accused of specific offending behaviour (classic school of thought).
Nonetheless, as already shown by Foucault (2008), the historical functioning of the penal systems is more complex than the opposition between theorical models of justice. The differences between the two schools are analysed by the author as an expression of the ambiguities inherent to the mixed character of the punishment system which combines the law with the so called 'normalising practices.' In a similar direction, Donzelot (2001) shows that the novelty brought about by the juvenile courts in France was the dissolution of the distinction between social welfare and criminal matters and, with that, the expansion of the judiciary's power to all corrective measures.
The ambiguity seems to also characterize the contemporary processes of transformation in the juvenile justice system. Although criticism to the rehabilitation ideal has undoubtedly played a key role in the contemporary transformations, scholars show that it remains an important principle in the juvenile justice, as a valued purpose for court workers (Ward & Kupchik, 2009), or combined with punishment in legitimizing the expansion of the system ("net-widening") with risk prevention and surveillance-oriented alternatives to incarceration (Case & Haines, 2021;Cate, 2016). Sallée (2017) shows how in France the rehabilitative philosophy has undergone a reconfiguration in the punitive context. With a renewed emphasis on the functions of confinement emerges a "rehabilitation under constraint" with individual responsabilization as a process and the offense at the centre. In the punitive framing, the individual responsibility of juveniles -associated by some authors with the context of neoliberalism alongside other new practices such restorative justice and risk management (Bailleau, 2002;Muncie, 2005Muncie, , 2008) -becomes one of the foundations of the system and is more often used to treat juveniles as adults than with to protect individual rights.

The Brazilian context
In Brazil, the emergence of juvenile justice was also discursively influenced by the positivist school of criminological thought. The criminological discourse of inequality was fundamental for the production of a new subject -the minor -and for the formulation of the legal treatment applied to their characteristics (See Alvarez, 1996). The ideals of the positivist school had a broad and positive reception amongst Brazilian jurists and intellectuals in the late nineteenth century and were decisive for the formulation of the first specialised legislation targeted at minors in the country: the 1927 Minors' Code. This piece of legislation abolished the discernment criterion From an offender-based to an offense-based justice: Changes… -device adopted in the first two Brazilian criminal codes to guide court decisions in cases of crimes committed by minors under the age of 14 10 -and incorporated legal provisions defended by the positivist school: the secret character of the judicial process and the abolition of the jury; individualisation of the penalty based on the study of the minor's physical, social, and moral characteristic; as well as indeterminacy of the sentence. The need for distinct legal-penal treatment for minors was justified with criminological knowledge of the positivist school of thought on the -biological, social, or moral -causes of criminal behaviour and the subsequent defence of a preventive treatment.
The conceptions that guided the way the Brazilian juvenile justice works received very few changes until as late as 1990. From an institutional point of view, an important milestone was the creation of the National Foundation for Minors' Welfare (FUNABEM) in 1964, which proposed national guidelines to reorganise official assistance to minors and set the creation of state foundations (FEBEMs) in charge of minors' institutionalisation (Paula, 2011, p. 36). Neither FUNABEM in 1964 nor the 1979 Minors' Code changed the focus on the causes of delinquency as a priority goal of this specialised justice. As with the previous legislation, the new code chose as its exclusive target audience a specific portion of the children population, designated by the category "minors in irregular conditions." The category included those under the age of 18 who were: deprived of essential conditions for subsistence, health, and education by their family; victims of abuse by parents or guardians; in moral danger; deprived of legal representation or assistance; with a history of misconduct due to family or community inadequacy; and perpetrators of a misdemeanour (Brasil, 1979, Art. 2).
The Brazilian juvenile justice system, therefore, has historically emerged with a focus on the identification and welfare of minors. This positivist approach centred in the figure of (delinquent and/or economically deprived) individuals lasted until the approval of the Statute of the Child and Adolescent (ECA) in 1990. In the wake of the approval of the 1988 Brazilian Constitution, the main change introduced by the new legislation is the abandonment of the minor category and the defence of children and adolescents rights (Schuch, 2005, p. 69-70). ECA also promoted a specialized juvenile justice system focused on special dispositions and institutions for adolescents in conflict with the law. As previously stated, detention has become a measure restricted to those accused of violent crime, recidivism, or repeated noncompliance with other measures imposed (ECA, Art. 122). The seriousness of the offense thus gains greater prominence as a criterion for judicial decisions and detention is now considered a way of holding the adolescent accountable for having committed a serious crime and no longer as a way of treating the different problems considered to be the causes for delinquency (See Méndez, 2006).
Albeit the operating logic of juvenile justice promoted by the ECA accompanies some of the transformations identified by the authors in other countries-especially the radical critique to the old model -it lacks a punitive or repressive direction. Even though it seems to be part of what Ward and Kupchik (2009) called the "accountability movement," it seems more directed to the "system accountability" for the protection of children's rights than to the "juvenile accountability" of the 'law and order' wave. The restriction to the use of detention is evidence of that.
The punitive logic appears, however, in the analyses produced by empirical research on the practical functioning of the system institutions. Some authors have demonstrated (See Schuch, 2005;Moreira, 2011;Adams, 2013;Vinuto, 2019), for example, that the emphasis on security procedures in the operating mode of detention centres has ever-closer proximity with prisons. Research on the procedures and decisions of the juvenile justice system has also signalled the importance of individual responsibility and of the seriousness of offense. Judicial decision-making (i) about whether or not to convict young defendant (Oliveira, 2017), (ii) to determine the period of detention (Almeida, 2014), or (iii) in the Superior Court of Justice (STJ) (Gutierrez, 2017) all emphasise the seriousness of the offense. In STJ decisions, the seriousness of the offense was decisive for the option to restrict procedural protection and expand criminal control.

Between 1990 and 2006: shifts in juvenile justice standards? The São Paulo context
Albeit ECA's enactment has changed the normative parameters of the treatment offered to adolescents in conflict with the law, the characteristics of detention centres have remained nearly the same. In the case of the state of São Paulo, the process of institutional change was driven by the crisis that hit FEBEM-SP in the late 1990s due to overcrowding of units, frequent cases of escapes and rebellions, and numerous allegations of torture committed by employees (Vicentin, 2005, p. 21;Paula, 2011, p. 67). In the 2000s, FEBEM-SP entered a restructuring process that involved the dismissal of employees accused of mistreatment 11 and the deactivation and the decentralisation of the centres. This process was intensified in 2006 when FEBEM-SP was replaced by the Foundation for Socio-Educational Support for Adolescents (Fundação Centro de Atendimento Socio-Educativo ao Adolescente, or simply CASA Foundation).
Thus, even though ECA was enacted in 1990, it took FEBEM-SP 16 years to be fully replaced by new CASA Foundation units. During this period, a peculiar situation took place in São Paulo: whilst normative principles no longer relied upon the From an offender-based to an offense-based justice: Changes… legal and social definition of minors, convicted adolescents were sent to FEBEM-SP centres -an institution that was developed exlusively to handle minors.
The coexistence of ECA principles with FEBEM-SP centres can be analysed particularly in the case of "Complexo do Tatuapé" -one of the most important and problematic centres for the confinement of adolescents in the twentieth century, where 17 centres operated within approximately 230,000m 2 , 1200 inmates, and 1,500 employees. This Complex was one of the most important symbols of FEBEM-SP, being frequently mentioned in the media for news related to human rights violations, especially with regard to the excessive number of inmates -Brazil even had to respond, in 2005, to a Precautionary Measure in the Inter-American Court of Human Rights precisely against "Complexo do Tatuapé", having been condemned and forced to take concrete measures to prevent further violations. 12 When FEBEM-SP was replaced by CASA Foundation in 2006, the Complex was deactivated and most of its buildings were symbolically destroyed, a public effort to demonstrate a definite rupture with the principles that guided previous legislations centred around minors. In its place, a public park, Parque do Belém, was built and is still open today in the borough of Tatuapé in São Paulo. Despite the destruction of these buildings, some facilities, such as CASA Foundation's administrative centres, remain untouched. This is the case of the Foundation's School for Professional Training and Qualification (EFCP), which hosts the Research and Documentation Centre (CPDoc). One of CPDoc's activities consists of the Adolescent Documentation Centre (NDA), a large collection of documents that aims to centralise, control, and update all information related to adolescents served by the Foundation (Alvarez et al., 2009). In particular, the NDA is responsible for managing all records and case files of every youth who was ever admitted to these units.
In 2007, right after the deactivation of the FEBEM-SP centres, the Centre for the Study of Violence of the University of São Paulo (NEV-USP) and CASA Foundation established a partnership through which NEV-USP researchers had access to a collection of 115,639 records and case files of every adolescent who was first admitted to FEBEM-SP between 1990 and 2006. This allowed the team to develop a series of research projects related to juvenile punishment and the institutional history of FEBEM-SP 13 (Alvarez et al., 2009;Salla and Alvarez, 2011). The team worked on two fronts: a quantitative and a qualitative one. First, 1,581 folders and case files were randomly selected to represent the population of 115,639 units, and a survey instrument was filled out (see Alvarez et al., 2009); second, an analytical form was developed for in-depth analysis of specific cases.
In this study, we rely upon the data representative of the population of records and case files of adolescents who were first admitted to a FEBEM-SP centre in the state of São Paulo between 1990 and 2006 to provide a quantitative assessment of the changes in the patterns in the period. In order to better understand our unit of analysis, a quick overview of the flow in the juvenile justice system in São Paulo is helpful.
12 See < http:// www. corte idh. or. cr/ docs/ medid as/ febem_ se_ 06. pdf > . 13 "Adolescents in conflict with the law: folders and records from 'Complexo do Tatuapé' (São Paulo/SP, 1990", submitted and approved by the MCT/CNPq 03/2008 cohort. The first contact an adolescent in conflict with the law has with the juvenile justice system happens though the police, when officers make arrests. The arrested adolescent is then taken to a police station, where an agent writes up a report and makes a decision on whether to open an investigation. Should an inquiry be opened, the adolescent is then either released or sent to a Primary Care Centre at a CASA Foundation unit. Soon after that, the adolescent to the Prosecution Office, where an informal hearing takes place. The prosecutor in charge, based both on the police report and on the informal hearing, then makes a decision on whether to release the adolescent or file a judicial representation, in which case the now defendants are sent to court -and potentially to a provisional detention centre until court day. A specialized judge then rules, and possible dispositions include pedagogical measures (such as detention), protective measures, judicial pardon, or even dismissal of the case altogether.
When the adolescent is institutionalised for the first time (either at the primary care centre, at the time of provisional detention, or upon conviction of a confinement disposition, an official institutional record is opened. It is essentially a series of documents that summarise the adolescents' progress within the institution: police reports, police inquiry, judicial representation presented by the Prosecution Service, and all other judicial and institutional documents. These documents are all filed at Complexo do Tatuapé. The unit of analysis of the empirical part of this study, therefore, refers to the population of adolescents who were institutionalised for the first time in a FEBEM-SP centre between 1990 and 2006 (See Alvarez et al., 2009;Vinuto, 2014;Oliveira, 2017Oliveira, , 2019.

This study methods and measures
From a sample of 1,581 records representing the population of adolescents who had their first institutionalisation in the state of São Paulo between 1990 and 2006, the data collected constitute a multilevel data set consisting of records (i.e., adolescents) nested within 2,312 court decisions -i.e., recidivist adolescents have multiple judicial decisions which add on their records. Albeit there are adolescents with up to 11 judicial decisions in the period, the average number of passages through the juvenile justice system is 1.64 -74% of adolescents had only one passage, and 16% had only two.
Data we analyse in this study consist of key information to the study of the juvenile justice system in São Paulo. Records' variables collected by the survey instrument include: adolescent's individual characteristics (gender, age, assigned colour, occupation, family references, education, drug use), collected from police reports; characteristics of the offense committed (the accused conduct, place of occurrence, date of occurrence), also collected from police reports; and sentence characteristics (date of sentence, sentence applied, amongst others), collected from court documents. In addition, to contemplate recidivism, blocks concerning the characteristics of the offense and the sentence were measured again for the adolescent's second passages through the juvenile justice system, and so on, successively. This study focuses on the relationship between the characteristics of the offense and the court From an offender-based to an offense-based justice: Changes… decision. In particular, we assess the extent to which this relationship has undergone considerable changes between the enactment of the ECA in 1990 and the collapse of FEBEM-SP units. The variables used are as the following:

Dependent variable
• Judicial decision: conviction to a detention disposition. Our focus is on sentencing patterns that suggest a punitive logic in the juvenile justice system, and being sent off to a detention centre for at least six months is the juvenile-justice-equivalent to imprisonment. 40% of court decisions included in our data set consist of the application of detention. This variable was coded as a binary indicator (1 = detention; 0 = other decisions).

Explanatory variables
• Date of court decision. Ranging from December 1990 to May 2009, 14 there are court rulings carried out throughout nearly two decades. In order to verify changes in sentencing patterns during this period, this variable was coded considering the year of the sentence. For a better interpretation of the coefficients, the initial year was considered zero, so that the distribution goes from zero to 19. Both the mean and the median of the decisions are equal to 11 (that is, 2001). • Offense. This is the criminal conduct of which the adolescent is accused, as described in the police report. The description in the police report did not always follow the standards of the Penal Code, so the categorisation of these responses was made according to more subjective criteria -see the Appendix for more details about the categorisation process. The main objective was to differentiate violent crime -such as murder, robbery, rape, and kidnapping -from non-violent offenses -such as theft, possession of stolen goods, embezzlement, amongst others. There are also records that clearly do not characterise offenses, such as loitering, squatting, and begging. The final distribution is as follows: violent offenses (52%), drug trafficking/use (12%), non-violent offenses (25%), nonviolations (5%), and records without information (6%). • Admission to the juvenile justice system. This is the variable that refers to the adolescent's criminal history (1 for first admission, 2 for second admission, and so forth). 68% of the analysed records refer to the adolescents' first ever contact with the juvenile justice system; 18% refer to their second time through the system, 7% on their third -all other rulings concern adolescent between the fourth and eleventh time going through the system. • Place of trial. Since different locations within the state of São Paulo may have different judicial organisations, it is important to control for where the ruling was issued. 54% of the decisions analysed took place in the capital's courthouse; 22% were carried out elsewhere in the state; and 11% in a municipality in the São Paulo Metropolitan Area other than the capital itself; 13% of cases did not contain any information on location.

Analytic strategy
As indicated, our work here has the main goal of analysing the sentencing patterns in the juvenile justice system and the possible transformations in such patterns over time. Given that the outcome variable follows a binomial distribution, the analytic strategy we adopt consists of binomial logistic regression models. Due to the multilevel characteristic of the data, models with fixed effects at the adolescents level were estimated. This implies that we are controlling for all individual characteristics that are constant at different passages through the juvenile justice system. More details below.
In order to assess shifts in decision-making patterns between 1990 and 2009, the sentencing date is an important explanatory variable. Is there an association between years and the odds of being sent off to a detention centre? During this period, did juvenile justice maintain the decision-making standards as suggested by the Minors' Code or, conversely, after the enactment of the ECA, did such standards change according to the predictions of the formulators of this new piece of legislation?
However, a positive association between years and the odds of a detention conviction does not necessarily indicate a change in sentencing patterns. It is possible, after all, that the standards of the sentencing process remained unchanged and that any changes to the proportion of times that detention was applied are due to other reasons -due to changes in criminal dynamics in the period, for instance. To assess whether there have been changes in the decision-making patterns, it is necessary to assess the extent to which correlates of court decisions have changed throughout the period.
In this sense, a key explanatory variable consists precisely of the seriousness of the offense. Unlike the minors' logic of previous legislations, the ECA doctrine limits the use of detention to cases of offenses that are either violent and/or pose serious threat to individuals. If the juvenile justice system has been adjusting to the dynamics of the criminal justice system over time, the weight given by court rulings to the seriousness of the offense might have changed -that is, it is possible that more and more judges rule aiming for the logic of proportionality between crime and punishment.
To test this hypothesis, interactive terms were included in the regression models. By interacting the offense categories with the year of the sentence, it is possible to assess the degree to which the association between each category and the odds of a detention conviction increased or decreased between 1990 and 2009. We hypothesise that the effects of violent offenses have increased when compared to effects of non-violent offenses.
Finally, two control variables were included. The adolescents' criminal history (i.e., was that the adolescent's first, second, third time in the system?) is an important variable because it also speaks to the logic of criminal justice institutions and because, without the inclusion of this variable, the effects of the date of the sentence are biased -considering that, due to the way in which data were collected, more recent dates tend to include more cases of recidivism. 15 In addition, the location where the court decision was made is also an important control variable, given that different locations may have different organisational characteristics and judicial patterns.
There are, of course, several other pieces of information (included in the data set or not) that could be sources of bias-variables such as gender, race, occupation, and family background. However, the inclusion of fixed effects at the adolescent level (which is possible because of multilevel modelling) implies that the only source of variation we are studying is the variation observed within each individual over time (i.e., recurrences) and not on the observed variation between individuals. Therefore, it is as if the models were already controlling for all individual characteristics that are temporally constant in the period considered -such as gender, race, occupation, family background, and any other unobserved sources of bias.

Results
Results of three binomial logistic regression models with fixed effects can be found in Table 1. Model 1 estimates the statistical effects of the date of trial, the seriousness of the offense, and the place of trial on the odds of a detention sentence. Model 2 includes criminal history as a covariate. Finally, model 3 includes interactive terms between seriousness of the offense and the date of trial.
Model 1 yields a positive association between year of trial and the odds of a detention sentence. Each year from 1990 to 2009 multiples the odds of a young defendant being sent off to a confinement unit (given that the coefficient is ̂ = 0.87 , which is equivalent to an odds ratio of exp ̂ = 2.39 ), controlling for the seriousness of the offense and the place of trial. As discussed earlier, the number of adolescents in custody has been increasing -even keeping the seriousness of the offense constant. However, this coefficient does not necessarily mean that sentencing patterns have changed.
In fact, the inclusion of the number of times that an adolescent has been in the system substantially modifies the coefficient for date of trial: it moves from ̂ = 0.87 in model 1 to ̂ = −1.20 in model 2; controlling for the seriousness of the offense, the place of trial, and the adolescent's criminal history at the time of the judicial decision, each year is actually associated with a drop in the odds of detention being applied by 70% (i.e., exp(−1.20) = 0.30)). According to model 2, an adolescent sued later in the period would have lowers odds of being sent off to a confinement institution than an adolescent who was accused of the exact same criminal conduct, at the same place, and with the same criminal history earlier in the period. The reason for this change is the inclusion of criminal history -recidivist defendants have substantively higher odds of detention ( ̂ = 2.84 , an increase of more than17 times for every additional passage through the system), and this is obviously correlated with date of trial because of the way in which data were collected. Yet, the number of adolescents in conflict with the law confined in detention centres is known to have increased (Brasil, 2019), so a possible explanation for the negative coefficient of date of trial in model might be related to a possible change in sentencing patterns in the juvenile justice system.
In addition to the temporal aspect, models 1 and 2 indicate a justice system that lies very close to that of criminal justice for adults. Controlling by the date and place of the sentence and their criminal history, adolescents accused of violent offenses are five times more likely to be detained than adolescents accused of non-violent offenses ( ̂ = 1.61 , which represents an odds ratio of exp ̂ = 5.00 ); and, as mentioned, each new passage through the juvenile justice system, regardless of the seriousness of the offense in question, multiplies the chances of the detention being applied by more than 17 times. The focus, however, should be model 3, which considers the interaction between the seriousness of the offense and the date of trial. Our goal is to assess the degree to which different categories of offenses have changed their weight in the decision-making process over the years. Whilst the impact of the year of the sentence is negative for non-violent offenses ( ̂ = −1.46 ), the weight assigned both to violent offenses ( ̂ = 0.27 ) and to drug trafficking or drug use by judges ( ̂ = 0.35 ) increased considerably compared to non-violent infringements. Figure 1 displays a graph in which these results are parsed out.
Model 3 shows that the date of trial is positively associated with larger probabilities of detention being applied in cases of violent crime. Even controlling for the number of times that an adolescent went through the juvenile justice system, the place of trial, and all other individual characteristics, the weight given to the seriousness of the offense increases considerably in the analysed period: every year increases the odds of a defendant accused of a violent crime being sent off to a detention centre by 31% in relation to non-violent arraignments. Even more surprisingly, offenses related to drug use or drug trafficking have also changed. The probability of detention in these cases increases substantially as time goes by: every year increases the odds of a defendant accused of some drug-related offense being sent off to a detention centre by 42%. This suggests that juvenile justice agents are increasingly more intolerant of drug-related arraignments; interestingly, the period we are analysing precedes the implementation of the 2006 act on drugs, 16 which presumably played a key role in mass incarceration in the adult penal system (Campos & Alvarez, 2017).
On the other hand, the graph shows the negative association between the date of trial and the application of detention for non-violent offenses: It is increasingly less common for adolescents to be institutionalised for less serious offenses. If, in the positivist logic that characterised previous legislation centred around the minor figure, the role of justice was to assess the individual offender, the juvenile justice system increasingly seems to be adopting a rationale that is closer to the adult criminal justice system -one in which a logic of proportionality no longer focuses on the offender but rather on the offense itself.
Results displayed here bring further evidence for the hypothesis that juvenile justice in São Paulo, since the enactment of the new legislation in 1990, has been moving closer to a classic logic that guides modern criminal justice systems -it has been moving away from a model of justice focused on the offender towards one focused on the offense.

3
From an offender-based to an offense-based justice: Changes…

Discussion
As detailed above, the international literature on recent transformations in the way juvenile justice works indicates a trend of transition from a model centred on the individual characteristics of the offender, in which intervention is applied as a way of dealing with the causes of the crime, towards a model centred on the offense, in which judicial dispositions are seen as a way of repressing serious offenses and punishing the offender. In this process, this form of specialised justice would lose some of its typical characteristics by moving closer to the criminal justice for adults. This approach, however, may have different meanings. The criminal justice model for adults can be used as a parameter to reduce discretionary decisions and expand the rights and procedural guarantees for adolescents. Furthermore, the concern with the characteristics of the offender and the interventions' goal of "treating" individuals never completely left the scene in penal policies for adults, and the decline of rehabilitation and resocialisation ideals is also recent within the criminal justice system.
We started this study asking the degree to which sentencing patterns had changed in the state of São Paulo between 1990 and 2006. This is an interesting period because a new legislation enacted in 1990 promoted an explicit rupture with previous legislations -rather than focusing on the minor figure, an umbrella identifier that collapses young offenders and children in precarious conditions, the new legislation focuses on offenses and offers pedagogical measures meant to rehabilitate adolescents in conflict with the law. Yet, convicted young offenders would still be sent off to FEBEM-SP units, an institution symbolically intertwined with the then obsolete treatment of minors. Between 1990 and 2006, a progressive legislation sought to promote measures to limit state intrusion upon the lives of adolescents, and old detention centres that were originally designed based on a tutelary system were adapted to account for the new legislative orientation -as such, principles of both the positivist and the classic criminological school of thought co-existed. Empirically, we assessed the degree to which the Brazilian juvenile justice system indeed started a pendular movement from an offender-based to an offense-based justice, with confinement measures symbolically representing the idea of punishment.
Results indicate that not only has the number of adolescents in conflict with the law confined in detention centres increased in the period, but crucially, the odds of being convicted and sent off to a FEBEM-SP unit varied considerably depending on the seriousness of the offense. As time went by, non-violent offenses were increasingly less punished, whereas violent crime was increasingly more associated with confinement dispositions. Interestingly, results also show that the juvenile justice system is increasingly tackling drug-related offenses more aggressively, with the odds of being sent off to a detention centre substantially shifting in the period. Confinement measures are thus more frequently being used as a response to specific types of crime. It is important to note that drug dealing is not a violent crime. The use of confinement measures as a response to this type of offense suggests that deprivation of freedom is being used as a way of controlling this type of criminality -something akin to the logic of the Brazilian criminal justice for adults -and not as a last resort in cases of violent offenses.

Methodological limitations
We of course recognise that this study has some limitations. First, the data that we analysed do not allow for causal inference. This is an observational study that does not seek to identify the causal effects of any phenomenon on court rulings. Panel data modelling is one of the great advantages of this study, given that we can remove time-constant confounding bias. However, these models are subject to bias generated by omitted variables that are not time-constant. The inclusion of the number of times adolescents went through the juvenile justice system and the place of sentence are important control variables, but it is precisely the possibility of the existence of other variables not included or even not observed that prevents us from inferring causality.
Also, the unit of analysis in this study is not ideal. As this is a study about sentencing patterns within the juvenile justice system, ideally it should encompass all judicial decisions taken in the period. By analysing institutional records filed at Complexo do Tatuapé, this study is limited to the population of adolescents who had their first passage through any FEBEM-SP centre in the period -that is, it excludes adolescents who went through the juvenile justice system (i.e., there was a court ruling), but who were never actually institutionalised (they were released by police officers, they were not provisionally detained, and/or they were not convicted with a detention disposition). However, other studies argue that this is an unusual profile that hardly reaches the time of the continuation hearing, when decisions are made (Oliveira, 2017). Additionally, the richness of this material is also demonstrated by the wide range of judicial dispositions found therein. In this sense, despite the limitation noted, we sustain that the validity of this study is not threatened.
In addition, we recognise that the period covered is somewhat dated. Important changes in decision-making standards may have taken place since the transition from FEBEM-SP to CASA Foundation units. Future research should look into this topic and update the results found here with more recent data.
Finally, while we demonstrate empirically that sentencing patterns in the Brazilian juvenile justice system did change over time, we are just inferring from such empirical patterns that the justice system operates through a pendular mechanism, moving from an offender-based approach of positive orientation to an offense-based approach of classic orientation; yet other theoretical mechanisms could be inferred from the same empirical patterns that we demonstrate. We caution the reader to critically assess the evidence, as ours is but one possible theoretically grounded explanation to the observed empirical patterns.

Final remarks
The goal of this article was to contribute to the analysis of the transformation process seen in the Brazilian juvenile justice system. Since the new legislation took over in 1990, adolescents in conflict with the law are held accountable; and confinement centres which were aimed at the treatment of minors are now exclusive for convicted young offenders. The analysis did not intend to measure the changes directly caused by the provisions of the legislation but rather the processes developed from the moment these provisions come into force. We sought to verify whether the direction of these changes coincided with the direction identified by researchers who investigated juvenile justice in other countries. Following the more general debate within the field of sociology of punishment on the expansion and strengthening of the punitive logic in penal policies as well as the functioning of the justice system, these authors identify a similar movement in juvenile justice as of the 1970s and 1980s. Drawing on data on the institutional transition of the socio-educational system in São Paulo after the approval of the new legislation, we verified the extent to which there were changes in the sentencing patterns related to conviction with detention, thus reinforcing a logic of punitivism.
With the analysis of the data presented, it is possible to claim that, after the approval of ECA, there was a change in the decision-making patterns of juvenile justice in the state of São Paulo: the seriousness of the offense has become more central for judicial decisions. In some ways, the emphasis on the seriousness of the crime as a criterion for the detention disposition is accounted for in the Statute of the Child Adolescent, which itself restricted institutionalisation to cases of offenses committed with the use of violence or serious threats. As indicated above, this approach in and of itself is not a sign that the punitive logic prevails in the type of response to offenses listed in the Statute. The establishment of this criterion for the application of institutional detention stems from the defence of the exceptionality of this measure by recognising its negative nature for the adolescent. Undoubtedly, by recognising such character, the function of holding adolescents accountable for their criminal conduct is highlighted, but the seriousness of the crime may come in at play not to expand the repression of certain crimes but, rather, as a means of limiting the deprivation of liberty.
In the case of the analysed data, however, the centrality of the seriousness of the offense seems to be associated with an increase in the punitive logic of the system of accountability of adolescents. Partly following the diagnosis in the international literature, the importance of the seriousness of the offense for the decision on the interventions seems to be part of a broader process. Although the data only provides an indicator of trends in juvenile justice, the empirical analysis confirms the aspects that have been identified in national surveys on the reinforcement of a repressive and security-based logic.
It is important to highlight that claiming that the seriousness of the offense is gaining centrality in the decision-making process does not mean that there are no mechanisms that favour a particular profile of adolescents. The typical attributes of "adolescents in conflict with the law" remain very close to those that defined minors: poor, black, male, residents of the peripheries of large urban centres. In this sense, it is possible that, in addition to being more punitive, juvenile justice in Brazil continues to reaffirm its selective and stigmatising features, already widely identified in the literature, both in the past and in the present.

Conflict of interests
The authors declare no competing interests associated with this manuscript.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.