The influence of social vulnerability and illicit drug use on recidivism of young offenders* A influência da vulnerabilidade social e o uso de drogas ilícitas sobre as reincidências de adolescentes infratores

The objective of this paper is to analyze factors related with the behavior of adolescents in Conflict with the Law through the study of the relationship between recidivism in youth detention centers (YDC) and their family context, social and neighborhoods context. In this paper we identify a negative correlation between a good performance on the Social Vulnerability Index (SVI) calculated in 2004, by districts in Porto Alegre, that means low social vulnerability and recidivism in YDC. The information used here was obtained from 138 Medical records at Fundação de Atendimento Sócio-Educativo (FASE). We propose a logistic regression model to understand how risk factors for delinquency in family context and social context correspond with the available statistics of adolescents in detention centers. In this capital city, an important factor to explain the recidivism in delinquency is the drug use of “marijuana-and-solvents” and “crack” when compared with the category of “denies or does not use drugs”. The results suggest that one should incorporate into the analysis the drug consumption between adolescents and their treatment in public or private health institutions. Additional work should be carried out not only in Porto Alegre but in other Brazilian cities. This work is original and useful for Law and Social Sciences research and public policies about Drugs, Delinquency and the Juvenile Court in Brazil.


Introduction
In this article, violence is measured by the infraction acts of adolescents in Porto Alegre (delinquency).
The city concentrated between 2002 and 2006 the largest absolute number of crimes in Rio Grande do Sul state (RS), with rates of 29, 21, 23, 24 and 19 homicides per 100,000 inhabitants, respectively.The rates of violencecorresponding to the sum of the number of threats, injuries and mistreatment -were between 2,200 and 2,500 cases per 100,000 inhabitants, per year, during the same period.However, the adolescent population is the subject of violence.In Schabbach4 the presence of gangs of children and adolescents is tackled by the analysis of violent crime for the 1990s and early 2000s in RS state.Nedel5 presents data from the 4 th Juizado da Infância e Juventude (JIJ), in Porto Alegre, from 2004 to 2006 observing the increase in the number of juvenile delinquents.Other studies in RS with information on juvenile offenders are Neto 6 , Neto 7 and Vasconcellos 8 .
According to Data Processing Company of the State of Rio Grande do Sul (PROCERGS) 9 , in the years 2005, 2006, 2007 and 2008 were consummated, respectively: 8, 12, 36 and 5 homicides against adolescents in Porto Alegre.For the same period and the city, the participation of adolescent offenders in homicides was: 25 (2005), 9 (2006), 9 (2007) and zero (2008).From 2007, there were more homicides committed against teenagers than those committed by them.In turn, the homicides attempts against adolescents reached 125 while there was a total of 64 attempts homicides by teenagers.None of these facts can be justified in any society.
The Vulnerability Index to Youth Violence (IVJ-V) 10

Theoretical considerations
In 1997, the National Council for Crime Prevention in Canada identified risk factors for the population under 18 years old.These risk factors were defined as "life experiences of young people that increase the chances of youth being victimized or of developing one or more behavioral problems" 13 .According to UNODC 14 , these and then proceed to estimate the multivariate model with selected variables.

Logistic regression model
In the logit model, the dependent variable is discrete (dummy) and may take binary values of 0 or 1. Departing from the traditional model of linear regression to estimate the expected value of (dependent variable), controlled by (explanatory variable or independent variables vector x): (1) There are no restrictions on the values of Equation (1), while assumes values between and .However, in the logistic regression model is predicted a probability value ( ) for the occurrence of an event, such that, it is necessary that the values are between 0 and 1.In the case of the logit regression model distribution's, functional form is given by Equation (2) below so that the values of the conditional mean of | are between 0 and 1: (2) The logit transformation of is in the following Equation (3)22 : The error term ( ) will not have normal distribution, but binary (HOSMER ; LEMESHOW) 23 .
To estimate the unknown parameters and in Equation ( 2) is used the method of Likelihood Maximum (LM) instead of Ordinary Least Squares (OLS).The LM method consists in to estimate a function that allows obtain the unknown parameters that maximize the probability of obtaining the results closer to the observed data.
Based on the results of the log likelihood function of Equation ( 4 (4)

Analytical procedure
Logistic regression is used to estimate the probability of recidivism ( ) (when more than one entry for an adolescent in a youth detention center -YDC-24 ), and in the right side of the equation are adolescents' personal information in the form of categorical explanatory variables.For example, for color-skin, if the base category or control category is color-skin White (1) and there are three categories, we have two dummies: 23 HOSMER, D.W.; LEMESHOW, S. Applied logistic regression.

Skin Skin_Parda Skin_Black
White (1) 0 0 Source: (The authors) We first evaluate the coefficients for univariate models and then proceed to estimate the multivariate model with selected variables.Then, we propose to estimate the model in Equation ( 5): (5) That is, we include variables of statistically significant categories, previously tested in contingency tables.
The group of three variables to be tested in the model is justified as: The rationale for minimizing the number of variables in the model is that the resultant model is more likely to be numerically stable, and is more easily generalized.
The more variables included in a model, the greater the estimated standard errors became, and the more depend the model becomes on the observed data.(HOSMER; LE-MESHOW25 ).

Personal and family atmosphere
The total number of measured aspects can be gathered in the dimensions of individual characteristics such as age when the delinquency act, gender, skin, aspiration, appearance, behavior, critical judgment, if suffers from some disease, if is influenced by others to commit the delinquency act, family involvement in crime and/ or in illicit drugs and alcohol, own drug use and alcohol consumption.Also, other characteristics of economic activity of the adolescent and his/her responsible (parents or responsible), as well as education, family structure (family composition, parents' marital status, other marriages of the parents), and delinquency act committed.
All these were explored based on the Individual Service Plan (PIA) -medical records-after authorization in ethics committee of our post graduate program.However, not all PIA were filled in the same way, and there are not digitalized but handwritten.

Results
The institutions consulted to obtain the informa-

Adolescents in Group 1
From 5,078 records of entries, from 2002 to 2008, the Temporary detention center Carlos Santos (CIPCS) and Center for Socio-Educational Care Female (CASEF)belonging to FASE -were extracted-by a convenience method-records of 1,251 adolescents (92% men).This data has the following information: race, age at which teenagers committed the delinquency act (AI), reason for the sentence of detention, the neighborhood district of family member or their parents and number of entries (recidivism means more than one entry).This is the group whose available information was complete.In the other records of entries (the others 3,827), there was missing information about the neighborhood or the reason for entry, or both, which is why these cases were not included, because without such information it was impractical to continue the study.vulnerability would be very high.From 0.5 to 0.7 SVI is low, which would indicate high social vulnerability.From 0.7 to 0.8 SVI is medium, indicating that the social vulnerability could be considered low.
The Table (1 are strongest, because they compound the SVI indicator. However, used them to measure the correlation with recidivism rate (RECID/ NGB) does not yield significant correlation coefficients.There is evidence, therefore, the suitability of the chosen indicator (SVI) everything else constant.
Table 1 -Matrix of zero-order correlations (Pearson) between SVI 2004 and the percentage of young offenders with recidivism, by neighborhood (RECID/ NGB)

Adolescents in Group 2
In order to verify the relationship between risk fac-       here is that illicit-drug use increases the probability of recidivism for the adolescents Group 2 at FASE.
) in the logistic regression, is feasible test the statistical significance of the estimated coefficients, calculate interval of confidence, and significance tests performed on the model (Wald test, Score test, and the likelihood ratio test).
tion and data analyzed in this study were: Judiciary of the State of Rio Grande do Sul through the 3rd Circuit's Juvenile Court for Children and Youth (JIJ) and the Central Court and the Bureau of Statistics and Register of Judges Activity (SERAJ), the Foundation for Social Welfare and Citizenship (FASC) of the city of Porto Alegre, through the Foundation for Social-Educational (FASE) Rio Grande do Sul.Of the total number of adolescents in conflict with the law, were chosen those that receive a judicial admission -and were turned off -at FASE, in Porto Alegre, from 2002 to 2008.The variable considered for measuring delinquency in adolescence was the readmission (recidivism) of adolescents in detention center, specifically, temporary detention center Carlos Santos (CIPCS) and Center for Socio-Educational Care Female (CASEF).
Figure (1) below illustrates, by age, the frequency of AI and recidivism in absolute values.Considering the range between 12 and 18 years old, the trend is increasing AI in relation to age.The AI has a peak at age 17, while the peak of recidivism is 16.After these ages, the values decrease, having in mind that when they are 18 years old, the youth is no longer under the jurisdiction of the Statute of the Child and Adolescent (ECA).The maximum age to stay at the Center of Detention is up to 21 years old, and the maximum time is up to three years 26 .

Figure 1 -
Figure 1 -Frequencies of delinquency act (AI) per age and recidivism per age in the group of 1,251 adolescents in detention centers (CIPCS and CASEF) 2002-2008.
) shows the zero-order correlation (Pearson's correlation).There is a moderate-low correlation equal to -0.35 (10% significance level) between the percentage of recidivism by OP sector with the SVI 2004.This means that the higher the SVI 2004, the lower social vulnerability and the lower the percentage of recidivism by neighborhood (RECID/NGB).At the level of neighborhoods, this result is consistent with the hypothesis that high social vulnerability is a risk factor for delinquency.The SVI 2004 correlations with the variables of percentage of the population with householder with less than four to eight years of study (4-8 STUDY) and the percentage of the population with householder up to two minimum monthly salary income (2 SMR) (in 2000) The author.Note: In parenthesis two-tailed test of statistical significance.*** P <0.01, ** p <0.05, *p <0.1.(P-value).Software SPSS Statistics 18.0.
tors and social or family context, was prepared a sample of 138 medical records of adolescents at CIPCS-CASEF, randomly selected.The sample size was calculated considering a percentage of judicial sentence for detentions in Porto Alegre ( ), to be representative of the population of juvenile offenders from 2002 to 2008.The total number of measured dimensions can be gathered as general characteristics, other features from the adolescent and from his/her parents (or family members) are their economic activity, as well as their education, family structure, habits and influence of the environment, among others (for example, physical health, mental health etc.).All those variables were explored in the Individual Service Plan (ISP) (from psychiatric evaluation at FASE) of 138 adolescents' records.
cal variables -vector x -having as a dependent variable the probability of recidivism (if not recidivism ) at CIPCS-CASEF.Only three of these categories were statistically significant.Interpreting OR (with the sign of the coefficients obtained) instead of the marginal effects, the category Skin_3 = BLACK increases the probability of recidivism 2.44 (144%) times in relation to the status of control (Skin_1 = WHITE) at 5% significance.The category Situation_at_Home_2 = Home STRUCTURED decreases the probability of recidivism 0.17 times (less 83%) in relation to the base category (Situation_at_Home_1 = Home in CONFLICT) at 10% statistical significance.The category Illicit_Drug_Use was significant for all cat-egories when the base category Illicit_Drug_Use_3 = CRACK, however, when shift to the base category Illic-it_Drug_Use_4 = no use or DENIES the interpretation of OR indicates increases of 1.06 and 2.70 times respectively in the probability of recidivism when Illicit_Drug_Use_2 = marijuana and other (solvents), significant at 10% and when Illicit_Drug_Use_3 = Crack, significant to 1%.For Illicit_Drug_Use_2 = marijuana and other (solvents) in contrast with Illicit_Drug_Use_4 = no use or DENIES there is an increased ratio of probabilities (odd-ratio) of 1.06/1.0(or 6%) of recidivism.And for Illicit_Drug_ Use_3 = Crack in contrast with Illicit_Drug_Use_4 = no use or DENIES there is a increased ratio of probabilities (odd-ratio) of 2.7/1.0 (or 170%) for recidivism.
High Social Vulnerability in the neighborhoods where adolescents in detention centersCIPCS or CASEF (2002 -2008) lived with their parents or family members in Porto Alegre, were risk factors for delinquency, as the initial hypothesis of this article.Ho-wever, there was no significant difference between living in a neighborhood or another, to return to recidivism during the studied period, from 2002 to 2008.This is due to the fact most of those responsible for adolescents lived in neighborhoods with high or very high social vulnerability, this discovery was made during the research.There is evidence of a negative correlation between a high value of IVS2004 for a neighborhood and recidivism: when the lower social vulnerability, the higher IVS2004 indicator and the lower the probability of recidivism.The way in which social interaction occurs at the neighborhood level was not detected in this study.Still, the risk factors for delinquency interact together rather than separately.For example, identifying other risk factors in Group 2 of the research -medical records of 138 adolescents -a variable with statistical significance to explain the probability of recidivism (Y = 1) was the illicit-drug use in the categories marijuana-and-solvents and crack, compared with the category of "non illicit-drug use or deny".We consider that the dimensions of individual personality and family context should be explored in studies of violence although in this paper, the results were not conclusive to affirm that: family structures of one kind or other, exerts influence on the behavior of delinquency (or not) of adolescents.The 138 medical records gathered, in part, life stories of teenagers with troubled lives of one or multiple forms: low education and/or difficulty to continue their studies, learning difficulties, deprivation, family conflicts or fragile structure (broken homes), relatives involved with illicit drugs (use and/or trafficking), crime and alcoholism and illicit drug use, parental abandonment, rejection, peer pressure and threats of traffickers, domestic violence, among others.Although the data sources and used bases are very useful, it is necessary to systematize the information of all adolescents at FASE thoroughly and carefully for further research in this line.For example, for a total of adolescents that were in detention centers CIPCS or CASEF, there was not systematized data about illicit-drug use.The information used here was obtained from 138 Medical records (with handwritten templates by Social Workers) at FASE.The sample was randomly selected and the sample size was calculated statistically.The same lack of information about the family or socioeconomic dimension of each of the adolescents, also limited the methodological implementation.However, the model highlighted the influence of one variable: illicit-drug use as explanatory for increase in recidivism probability.Yet, there is not enough information about the direct causality between the consumption of illicit-drugs and delinquency.This issue can be studied in further research in Brazil, as it has been done in other countries in the world.What is said in the results -by the Brazilian Forum on Public Safety (FBSP) in partnership with the State System of Data 6 NETO, L.A.M. Internações no Centro da Juventude de Santo Ângelo: diferenças e peculiaridades com as demais unidades de internamento da FEBEM, localizadas no Rio Grande do Sul.2001.128 f.Monografia de especialização (Especialista em Métodos Estatísticos) -Programa de Pós-graduação em Métodos Estatísticos, Universidade Regional Integrada do Alto Uruguai e das Missões, Santo Ângelo, 2001. 7NETO, L.A.M. Características dos adolescentes infratores internados no centro de atendimento sócio-educativo regional de Santo Ângelo e modelagem do fluxo de ingressos na fundação de atendimento sócio-educativo do estado do Rio Grande do Sul.2003.190 f.Dissertação (Mestrado em Ciências no Domínio da Modelagem Matemática) -Programa de Pós-graduação em Modelagem Matemática, Depto.Física Estatística e Matemática.(DeFEM), Depto.de Tecnologia (De-TEC), Universidade Regional do Noroeste do Estado do Rio Grande do Sul, Ijuí, 2003. 8 10The IVJ-V was calculated for the population 12 to 29 years in 266 Brazilian cities with more than 100.000inhabitants and consists of a weighted average that includes the following indicators: homicide mortality indicator, an indicator of mortality from traffic accidents, indicator of school attendance and employment, an indicator of poverty and inequality.BRASIL.Ministério da Justiça.Projeto Juventude e Prevenção da violência.Available in: <http.//www.forumseguranca.org.br/institucional/wp-content/uploads/2009/11/relatorio_ pjpv_2009.pdf>.Access in: mar.2010.
The influence of social vulnerability and illicit drug use on recidivism of young offenders

Table ( 2
) brings the test's results for seven categori- The influence of social vulnerability and illicit drug use on recidivism of young offenders