Criminal Behavior and Mental Health Problems among Adolescents: A Cross-sectional Study and Description of Prevention Policy in Sweden

The present study investigates the association between mental health problems and criminal behavior among adolescents in Sweden. Community crime prevention in a Swedish context is also discussed. Every two years, pupils from schools in Stockholm answer the Stockholm School Survey with questions and statements about their social situation, alcohol and drug use, attitudes, school climate, school grades and criminal behavior. Data collected from pupils who answered the survey in 2014, 2018 and 2020 form the basis of this study. A significant association between mental health problems and criminal behavior was shown, even after controlling for factors suggested in international literature. Mental health problems were shown to be a strong explanation for criminal behavior among adolescents.


I. Introduction
Juvenile delinquency generally has severe consequences both for the individual and for society. In general, various studies have shown that many but not all crimes are related to mental health problems. 1 This study analyses mental health problems and criminal behavior among young adolescents in Sweden. As an introduction, and to present a backdrop for discussion of study results and conclusions, an historical overview of criminal policy for young offenders in Sweden is given first.
The way states handle juvenile offenders and the connection between criminality and mental health problem differs both between states and over time. In the case of Sweden, for a long period of time, at the least from the 1950s and onward, the main aim in criminal policy concerning young offenders has been to keep children and youths out of the prison system, as far as possible. The legislature's advocacy for this policy is due to the view that young offenders aren't sufficiently psychologically mature to be deemed completely responsible for their actions. Therefore, this view implies that young offenders, to refrain from further criminality, likely need care or treatment and support from social services rather than facing punitive measures. A consequence is that young offenders' actions are less censurable, and the penal value is reduced or waived of prosecution. 2 From a Nordic perspective, i.e. that of the Scandinavian countries, this kind of policy is not unique. However, in comparison to other European countries it is. In relation to the Swedish standpoint, Jordan Pratt argues that this position in Swedish criminal policies emerges from the "culture of equality" that existed in the Nordic countries, which is embedded in their social fabrics through the universalism of the Scandinavian welfare state. 3 A motive for this culture can be traced to long-lasting social democratic rule in Sweden. Or, as Gösta Esping-Andersen's classical typology of welfare regimes states, Sweden and the other Nordic countries are historically to be seen as social democratic welfare regimes. 4 Since the 1990s, the domination of social democratic parties in the Nordic countries has been broken, and several conservative and/or liberal parties have had a strong influence on national politics. Although some changes, i.e. increased penalties for violent crimes, have been introduced in the last decade, the main policy in Swedish criminal policy on young offenders still is to keep the majority of criminal youths out of the prison system.
In Sweden, the age of criminal liability is fifteen years, which it has been since the beginning of the twentieth century. 5 Today, several procedural and penal laws exist which state that young offenders, foremost those fifteen to seventeen years of age, but also those eighteen to twenty years old, should be handled differently than adults. Responsibility for societal reaction towards children, under fifteen years lies solely with the municipal social services. For youths fifteen to seventeen years old, social services and the criminal justice system have a joint responsibility. For young people eighteen to twenty years old, the principal responsibility for societal reaction lies with the criminal justice system. 6 The "stay out of prison policy" entails that juvenile delinquents, depending on age, receive either care within social services, healthcare treatment, or punitive measuresor a combination of all three. Due to this predominant view, Swedish authorities, mainly through the municipal social services and the police, use various preventive measures to hinder youth from Criminal Behavior and Mental Health Problems among Adolescents getting involved in criminality at all, or to encourage juvenile delinquents to refrain from further criminality.
In a similar manner, the view of children and young people with deviant behaviors, deriving from criminality and/or mental illness has varied over time, both in Sweden and in other Western countries. Ulrika Levander and Lina Sturfelt describe the overall historical development of Swedish views on mental illness and crime in three main stages. 7 Firstly, the social medicine view prevailed between 1902 and 1950. This view entailed that individuals with psychosocial problems were seen as morbid and perhaps contagious, i.e. that society should be protected from those considered deviants. This meant that psychiatric or social institutions grew in number, to care for youth and adults with deviant behavior. Secondly, the period between 1950 and 1989 was characterized by psychodynamic object relations thinking. Attachment to peers and careers was emphasized as decisive for the child's development. This led to the view that societal functions, i.e. municipal social services and child psychiatry should be involved in measures to counteract deviancy to a much higher degree, where they cooperate in the care of children and young people with social and psychological problems. Unfortunately, the legislature did not specify which party should have the main responsibility for implementing this new deal in treatment cooperation. This ambiguity led to circumstances where not all children and youth with social and psychological problems were given adequate care or treatment. Thirdly, from 1990 onwards, the view of mental illness changed, in Sweden and many other countries, to be viewed as an extensive public health problem. That is, the causes of mental health problems are often seen as related to the surrounding environment. With this view, a strong emphasis on preventive efforts against both mental illness and social problems has emerged. 8 Developments throughout the twentieth century have, finally, meant that the individual's responsibility for his/her own mental health is reduced and that society today takes a greater responsibility for preventing mental illness among youths and adults.
We have so far briefly reviewed the developments in a Swedish context, but what do we really know about the connection between mental health and criminality? Previous research has shown that such a connection exists. For instance, a Swedish study of a representative sample of 411 adult inmates found that inmates could be described as a marginalized group with higher prevalence of social and mental health problems compared to the general population. 9 A similar Canadian study examined the link between psychiatric symptoms and criminal behavior in everyone living in Manitoba who had been in contact with the Prison and Probation Service or who had been a victim of crime. 10 Individuals in the general population in the same area, not previously sentenced, were used as the control group. It turned out that both those arrested for crime and victims of crime had a significantly higher prevalence of mental health problems compared to the general population. An overall connection between social and mental health problems and criminality seems to exist, and the same seems to be true H. Källmen et al.
160 regarding juvenile offenders. Several explanations have been proposed to understand how juvenile mental health issues play a part in trajectories of delinquency.
Sensation seeking seems to be a key variable in delinquency. For example, a Turkish study investigated the relationship between sensation seeking and attitudes towards violence in youth. 11 Participants were 393 boys and girls between twelve and fifteen years old. The results showed a strong significant relationship between sensation seeking and a positive attitude towards violence. With the aim of investigating a possible relationship between crime and sensation seeking and impulsivity, respectively, Armstrong et al. surveyed 872 university students with a mean age of 20.35 years. 12 After controlling for callousness, egocentricity, and Machiavellianism (manipulativeness, indifference to morality), both sensation seeking and impulsivity were significantly associated with criminal behavior. However, from a theoretical perspective, sensationseeking and anti-social behavior can be perceived as normative for adolescents, a phenomenon that can be explained by an interaction between a maturity gap (biological vs. social) and the social peer context. As proposed by Terrie E. Moffitt, individuals whose anti-social behavior is reinforced and who mature early physically can thus exhibit delinquent behavior in adolescence, but only a small proportion continue to be delinquent as adults. 13 School can significantly influence, i.e. reduce or hinder, the effects of the social context on incipient delinquent behavior. In a study on 25,850 students fifteen to eighteen years old, in lower and upper secondary school in Stockholm, Sweden, the associations between crime and family factors, as well as the school climate were studied. 14 In this study it is suggested that unfavorable family factors were positively associated with selfreported crime and that a good school climate was negatively associated with crime. The perception of schoolwork as meaningful was interpreted as a factor that could compensate for poor family conditions. At the individual level, crime among young people is also associated with poorer school performance and earlier interruption of studies.
Juvenile delinquency can be seen as a group phenomenon, where peers have a great deal of influence. It has been suggested that juvenile delinquency is socially learned through psychological mechanisms such as imitation and reinforcement and that close friends have a particularly large influence. Jinho Kim and Jason M. Fletcher studied whether a wider group of peers like classmates or schoolmates has an impact on young people's criminal behavior, based on data from a longitudinal study of juvenile delinquency and health status conducted in Wisconsin, USA. 15 Participants were 16,912 young people in school in grades seven to twelve. Juvenile delinquency was defined as having committed any of the following crimes in the past year: stealing something worth $50 or more, committing an armed robbery, breaking into property, or selling drugs. The results showed a slight association between the proportion of peers with a criminal background and reports of one's own criminal behavior. An increase in the proportion of peers with a criminal background by five percent was associated with self-reports of Criminal Behavior and Mental Health Problems among Adolescents increases in one's own criminal behavior by three percent. The cumulative risk model in a criminological context assesses risk factors for juvenile delinquency. A recently published retrospective study evaluated cumulative risks in a cohort of 3 414 men in late adolescence, all identified in the 1986 birth cohort and living in northern Finland. 16 Known risk factors were measured by midwives during the mothers' pregnancy, questionnaires about risk factors during childhood, for ages seven to eight, were mailed to the children's parents and head teachers in school. Reports of crimes between the ages of fifteen and twenty were taken from the national criminal registry. Crime was defined as property crime, driving while intoxicated, drug-related crime, and violent crime. The study demonstrated strong links between risk factors present during pregnancy and childhood, and crime in late adolescence. Gregory M. Zimmerman and Steven Messner investigated violent crime among adolescents between the ages of ten and twenty-two in Chicago, USA. 17 They showed that the probability of a person committing a violent crime was significantly associated with the proportion of peers who commit violent crimes, and that women were affected more than men. They also found that growing up with one biological parent compared to two biological parents increased the likelihood of committing violent crimes. The conclusion was that in problem residential areas, there is less difference between the sexes in crime than in more stable residential areas. This is due to the fact that young women compared to young men are more affected by lack of support from parents and peers.
Similar results are also shown in Sweden. According to a national survey from the Swedish National Council for Crime Prevention [Brå] including 5 050 pupils in ninth grade in lower secondary school half, of the pupils state they had committed a crime (theft, violent offence, vandalism, narcotic offence) on at least one occasion during the last year. 18 The differences between girls and boys in criminal propensity are quite small; however, boys show an overrepresentation for more violent offences and vandalism. Similar to Zimmerman and Messner's results, Swedish pupils with divorced parents also more commonly state that they participated in crime. 19 Similar overrepresentation in crime can also be seen for pupils from families with limited financial resources. 20 Regarding associations between mental health among adolescents and criminal behavior later in life, Mark D. Anderson et al. followed a cohort consisting of pupils in grades seven through twelve in 132 schools in the US, focusing on depression. 21 The authors adjusted for factors that correlate with both depression and criminal behavior. They were able to demonstrate a significant association between depression among adolescents and property crimes in individuals aged twenty-five to thirty. With the aim of studying the long-term effects of children and adolescents' depressive symptoms, William E. Copeland et al. conducted a study of 1,420 adolescents in North Carolina. 22 Repeated measurements were conducted annually between the ages of nine and sixteen, and again at nineteen, twenty-one, twenty-five, and thirty years. The results showed that depression in H. Källmen et al.
children and adolescents is a significant predictor of criminal behavior among young adults. When the analysis was adjusted for criminal behavior in adolescence, the association became insignificant. This can be interpreted to show that the occurrence of criminal behavior in adolescence, in combination with depression, predicts criminal behavior in young adults. Another longitudinal study with four measuring points between 1994 and 2006 targeted children and adolescents between the ages of six and seventeen who attended school in Zurich, Switzerland. 23 The aim was to investigate predictors of crime as adults, i.e. fifteen years later. The results showed that traffic offenses and drunk driving, but also violent and drug-related offenses were the most common offenses for these adults. It was concluded that low socio-economic status, problematic alcohol consumption and outspoken behavior in adolescence are predictors for criminal activity in the twenty-five to thirty age group.
Looking at the combination of mental health problems and social problems, a literature review showed that adolescents' encounters with the prison service often are preceded by both mental health problems and social problems. 24 In a longitudinal study of a birth cohort, a significant relationship was shown between low self-esteem in adolescence and problems with criminality as an adult, after controlling for depression, gender, and socio-economic status. 25 A study of 164 adolescents aged eleven to seventeen (and their parents), convicted of crime who were undergoing treatment for mental illness in two North eastern cities in the US. 26 The relationship between psychopathology and the severity of the crime was examined. It was shown that serious crimes and violent crimes were associated with outspoken, externalizing behavior while internalized behaviors such as anxiety and depression were evenly distributed along a continuum of crime severity.
In Sweden, as in the other Nordic countries, the municipal Social Welfare Boards have the ultimate responsibility for community efforts to hinder children and adolescents to develop criminal behavior. This implies that Swedish social services use both selective and indicated prevention measures targeting groups or individual youths who are at risk of developing, or have already developed, such behavior. In Sweden such preventive efforts are organized in cooperation with other authorities, preferably schools and the police. By the Swedish Social Services Act, all preventive measures to children and youth at risk, are primarily provided on a voluntary basis. 27 Mandatory indicated measures for children and adolescents are only possible under specific circumstances. That is if the necessary prerequisites of the Care of Young Persons (Special Provisions) Act are met. 28 A decisive prerequisite is that there needs to be a significant risk of harm to the child's future health or development. This law applies under two different conditions, either if the child needs care due to the situation in the family home (e.g. abuse), or due to the child's/adolescent's own behavior (e.g. criminality or misuse of drugs) and the situation is considered harmful for his/her future health or wellbeing. If the prerequisites are met, after an application by social services and a decision by an Criminal Behavior and Mental Health Problems among Adolescents administrative court, the child/adolescent is usually placed in foster care or residential care. Such placement could be combined with outpatient care or treatment. However, in the case of the youth's own criminality it should be said that the Swedish "stay out of prison policy" entails that juvenile delinquents can be surrendered to care within social services or sentenced by criminal court to special sanctions reserved for youths (specifically fifteen to seventeen years old, but in special cases up to twenty years of age). These sanctions are multiple, but the most commonly used are those of the Penal Code's section Care of Young Persons. 29 The authority for enforcement of such sanction are the municipal social services. During the enforcement the youth is committed to specific treatment and might also be placed in foster care. In the case of more serious offenses, for instance murder or rape, such a sanction will not be deemed appropriate by a criminal court. The youth, fifteen to seventeen years of age, will likely be sentenced to the sanctions indicated in the section Institutional Care of Young Persons of the Penal Code. 30 This specific sanction implies placement in a special state institution for up to four years. Legally, it is possible for a criminal court to convict a young offender over fifteen years of age to prison, but according to the "stay out of prison policy" this sanction is not applied. All specific sanctions for juvenile offenders have the purpose of hindering the youth to relapse in further criminality. Legally, and ethically, both the Act of Special Provisions and the specific sanctions for youths' rest on Article 5d of the European Convention on Human Rights [ECHR]. 31 All measures in the legal process and enforcement of sanctions shall also comply to the UN Convention on the Rights of the Child [CRC]. 32 Since the 90s, several governmental inquiries, with directives to review the sanction system for young offenders, have been undertaken. Despite some minor legal revisions, the "stay out of prison" policy is intact. The legislature's view still is that children/adolescents who commit offenses often are socially disadvantaged and in need for help and support rather than punitive sanctions. In reference to the CRC, the Swedish government has stated that society's reaction to youth criminality must be directed at rehabilitation and measures to combat risk factors that contribute to the child's criminality and may lead to reoffending. 33 However, since 2021, a shift in criminal policy on youth offenders can be seen. This shift has taken place in response to rising gang criminality in Sweden. For example, the sanctions specified in the Penal Code section Care of Young Persons are today to a higher degree directed towards young offenders fifteen to seventeen years of age and only in special cases (rarely) for offenders up to twenty years old. 34 The effect of the shift in criminal policy has been that for juveniles eighteen years and over, who have committed more serious and violent offences, prison sentences should be adjudicated. So, the "stay out of prison" policy has somewhat altered. Today, specifically juvenile offenders fifteen to seventeen years of age are to be treated differently, while this is not to the same extend the case for those eighteen to twenty years old. This specifically concerns the criminal courts' choice of penalty for more serious and violent offences.
The first aim of this study is to evaluate the association between juvenile delinquency and mental health problems, such as depression and anxiety, while also considering factors such as socio-economic status, grades in school, sensation seeking, and being friends with criminal peers. After identifying other factors associated with delinquency the study's second aim is to discuss how juvenile delinquency problems are, and how they can be, addressed at selective and indicated levels in a Swedish context.

Participants
The Stockholm school survey is completed every other year by pupils in lower secondary school (grade nine, the last compulsory year in the Swedish school system) and upper secondary school (grade eleven, the second of the three non-compulsory years in the Swedish school system). The survey is mandatory for public schools, but voluntary for private schools. The purpose of the survey is to inform the decision making of local authorities, and ultimately with the purpose of improving pupils wellbeing and, in the long run, their school performance. The questions relate to life circumstances, schoolwork, bullying, drug use, health, and crime. Noncompleters are those who for some reason were absent from school when the survey was administered. Available data consist of responses to the survey in 2014, 2018, and 2020. Response rates on school level were 76% in 2014, 73% 2018 and 72% in 2020. In 2014, 5,235 boys and 5,761 girls responded, in 2018 there were 5,017 boys and 5,211 girls, and in 2020 there were 5,633 boys and 5,865 girls who responded to the survey. The responses at each measurement occasion are compiled into a single data file and analyzed simultaneously.

Procedure
Pupils completed the questionnaire during a school lesson and then placed it in a sealed envelope before handing it to their teacher. The pupil has the entire lesson (about forty minutes) to complete the questionnaire. Pupils are informed that it is voluntary to participate and that they are free to withdraw their participation at any time without consequences.
Pupils also receive information that the Origo Group, a private evaluation company, is responsible for the collection of the data on behalf of the City of Stockholm.

Gender Identity
For pupils stating boy, this was coded 1, and for those stating girl, this was coded 2.

Criminal Behavior
The Stockholm school survey contains nineteen questions about criminal behavior, from trivial crimes such as shoplifting and free riding on the subway to serious crimes such as burglary and unlawful coercion. In order to divide the crimes into groups, an exploratory principal axis factor analysis was performed.

Criminal Behavior and Mental Health Problems among Adolescents
The analysis showed that the nineteen questions could be divided into three latent factors with eigenvalue over one which together explained 52.4% of the co-variance. One of the factors that alone explain approximately 8.6% of the co-variance contains six questions about dealing with stolen goods and violent crimes. They were: "Have you during the latest year … 1. Stolen a bicycle? 2. Sold something you know was stolen? 3. Bought something you know was stolen? 4. Burglarized a car, store, kiosk or other dwelling? 5. With intent hit somebody so that you believe or know he or she needed to visit hospital? 6. Carried a weapon, (knife or knuckle-duster)?" In this study, those questions were used to define criminal behavior and constituted an index with high internal consistency, Cronbach's alpha 0.81. Respondents who did not report any of those behaviors were coded as 0, and respondents who reported at least one of those behaviors once during previous year were coded as 1.

Mental Health Problems
Occurrence of mental health problems were assessed by using a modified version of the Psychosomatic Problem Scale (PSP-Scale). 35 The modification consisted of deletion of items asking for difficulties to concentrate and feeling giddy and adding an item about feeling great to be alive. In this study we used seven different symptoms or problems, such as headaches, depression, feeling fear, stomach problems, difficulty sleeping, thinking that it is great to be alive (coded negatively as seldom or any single time), and poor appetite. A principal axis factor analysis showed that a single factor with eigenvalue above 1 could be extracted and those seven items all loaded on this factor. Five items loaded above 0.5 and items about headache and feeling fear loaded lower. The answers of pupils who responded (on a five-point scale) that these problems occurred at least once a week functioned as indicators of possible mental health problems. Internal consistency was fair, Cronbach's alpha 0.69. By adding up these problem areas, an index was created that was scored from zero to seven on mental health symptoms. Those who scored between zero and four symptoms at least weekly were considered to have a low indication of mental health problems, while those who had between five and seven symptoms weekly were coded as probably having mental health problems.

Socio-economic Status
Socio-economic status was assessed with three questions about the mother's education, the father's education and the pupil's allowance, i.e. the money per month the pupil receives for their own consumption. If the mother had a university education, 1 was coded; if not, 0. If the father had a university education, 1 was coded; if not, 0. If the pupil received more than SEK 1,000, approximately $120, 1 was coded, if not, 0.

Sensation Seeking
The Stockholm school survey contains twenty-one statements about attitudes and behaviors in different situations. In order to divide the H. Källmen et al.
statements into categories, an explorative principal axis factor analysis resulted in five factors, that together explained 53.4% of the covariance. The main factor explaining 27.2% of the co-variance consisted of six items with sensation seeking content. The respondents were asked to estimate on a four-point scale to what extent the statement was true for them. The responses were given on a four-point scale of very bad, bad, good, and very good, coded 1-4. The six statements are: "1. I like to do exciting and dangerous things even if it is forbidden. 2. I am often outdoors together with friends on nights. 3. I ignore rules that prevent me from doing what I want to do. 4. I look at myself as a rather impulsive person. 5. I want to be there where exiting things happens. 6. I do stupid things even if they are little dangerous." These six items constitute an index with high internal consistency, Cronbach's alpha 0.81. The index was scored between one and twenty-four and dichotomized using a median-split. Participants scoring from lowest to median were coded as 0 and those scoring above median were coded as 1.

Grades in School
Grades in school in the subjects Swedish, English and mathematics were self-reported and measured on a seven-point scale where 1 was coded if the teacher had not been able to assign a grade for the pupil, F [failed] was coded 2, E [pass] was coded 3, D was coded 4, C as 5, B as 6 and A [top grade] was coded 7. Grades A, B, and C comprised more than half of the pupils in each subject. In order to dichotomize, A, B, and C were considered good grades and coded 1, the other grades were coded 0.

Criminal Friends
The proportion of criminal peers was measured by a single question: "How many of your peers inside and outside school have committed burglary, or robbed someone, or stolen a car?" The answers were given on a four-point scale from no one, a few, about half or most. The answers were dichotomized, and no one or a few were coded with 0 and about half and most were coded as 1.

Ethical Considerations
As the data is anonymous, our study was exempted from ethical approval according to an earlier decision of the Ethical Review Board in Stockholm Dnr 2010-241 31-5.

Statistical Analyses
All analyses were performed using SPSS v. 26.0. Descriptive statistics for the proportions of girls and boys who reported criminal behavior are presented. Additionally, we investigated the latent factors behind criminal behavior and sensation-seeking attitudes by conducting principal axis factorings. We also evaluated the gender-specific associations between mental health problems and criminal behavior when adjusting for socio-economic variables, grades in mathematics, Swedish, and English, sensation-seeking attitudes, and the number of criminal friends in a multivariable logistic regression analysis.

III. Results
The results showed that boys reported criminal behavior during the previous year at a higher prevalence (20.1%) than girls (8.8%). Table 1 shows the relationship between mental health problems and crime, defined as dealing with stolen goods and violent crime. The regression analysis has been adjusted for socio-economics, school performance, sensation-seeking attitude and criminal peers. After adjusting for relevant factors, the odds for criminal behavior almost doubled for adolescents with mental health problems compared with those without. The results show that if more than half of the peers are criminals, the risk for reporting one's own criminal activities increases almost six times for girls and almost eight times for boys. A sensation-seeking attitude was also strongly associated with reports of one's own criminal activity. Table 1 also shows that good socio-economic conditions and good school performance are negatively associated with reports of own criminality.
Interestingly, the results show that parents' education affects boys and girls differently. If the father has a university education, it is significantly negatively associated with reports of own criminality among boys, but not for girls. However, if the mother has a university education, a significant negative association to reports of one's own criminal activity among both boys and girls is the outcome. Receiving more than 1,000 SEK (about $120) for one's own spending every month is negatively associated with criminal behavior among girls, but not among boys. A significant negative association between grades in mathematics and reported criminality is shown among boys but not among girls. Table 1. Results from a logistic regression analysis (p-value, odds ratio, and 95% confidence interval) of the association between mental health problems and crimes of violence and those dealing with stolen goods in the previous year, by adjusting for socio-demography, grades in school, criminal friends, and sensation-seeking attitudes.

IV. Discussion
The first aim of this study was to evaluate the association between self-reported juvenile criminality and mental health problems, such as depression and anxiety, when also considering factors such as socioeconomic status, grades in school, sensation seeking and being friends with criminal peers. After identifying other factors associated with delinquency, the second aim was to discuss the nature of juvenile delinquency and how it can be addressed at selective and indicated levels in a Swedish context. The study results from our Swedish sample show that, after controlling for socio-economic status, school grades, sensation-seeking attitudes and criminal peers, the probability of reporting one's own criminal activities was almost twice as high among those who also reported mental health problems, compared with those who did not.
Both similarities and differences between boys and girls were identified regarding explanatory factors for crime. Having criminal peers, a sensation-seeking attitude, and mental health problems were the three strongest explanatory factors for criminality in both genders. Good grades in Swedish and English and maternal university education were negatively associated with criminal behavior for both boys and girls. However, good grades in mathematics seem to function as a protective factor for boys, but not for girls. This is interesting, since good grades generally mean high attendance in school, i.e. staying out of trouble, as Julia Sandahl has previously shown. 36 There were also other differences between the sexes. Paternal university education was significantly associated with less criminality among boys, but not among girls. Having more than SEK 1,000 per month for pupils' own use was significantly associated with less crime among girls, but not among boys. A prudent interpretation could be that boys, to a higher degree than girls in these age groups, are more sensation-seeking. If so, regardless of allowances for their own consumption, some boys are involved to higher extent in petty crimes, just for the sensation.
The results of this study confirmed the association between mental health problems and criminality shown in previous studies. 37 A recent study showed a positive association between adolescent mental health problems and a poor school climate, and also with low grades in Swedish. 38 Similarly, it has been shown previously that good school climate was a protective factor against offending among adolescents. 39 As expected, the present study showed a significant negative association between criminality and school grades, foremost in Swedish and English, where low grades increased the odds of criminal behavior between 20 and 40%. Consistent with earlier studies, this study demonstrated an association between criminal behavior and a high proportion of criminal peers. 40 This confirmed the findings of Moffitt, who proposed that antisocial behavior among adolescents, Criminal Behavior and Mental Health Problems among Adolescents including offending, is learned through mimicking adequate role models. 41 The present study showed that good socio-economic conditions are negatively associated with pupils' criminality, a finding previously confirmed by Marcel Aebi et al. 42 This finding also supports the Swedish standpoint that children/adolescents who commit offenses often are socially disadvantaged and are therefore in need of help and support. 43 It might be said that the Swedish "system" to help young offenders, particularly those fifteen to seventeen years of ageforemost within the social services seems to rest on some scientific evidence. If improved socio-economic conditions do lead to non-relapse in crime for a young offender, it should be seen as a benefit both for society and the juvenile offender him/herself.
Although paternal university education showed a significant negative association with offending among boys, this did not hold true for girls. One way of interpreting this finding is that it might have to do with the father's role as a model for identification among boys. The opposite was shown for allowances (above 1,000 SEK/month) for own use. This was negatively associated to criminality for girls, but not for boys, a finding that is possibly consistent with the proposal of Zimmerman and Messner that girls from socio-economically weaker families are more likely to show offending. 44 The level of allowances is a question for parents. However, the ability to give higher allowances may be limited if a family's income is low. A societal economic contribution, via the social services, directly to young people from low-income families, could possibly function as a preventive measure. Such contributions, i.e. "extra money" for social activities, has been shown to have good effects for adults with serious mental health issues. The persons' sense of self increased through their experience of, for example, agency, recognition, and security when receiving money slightly over their normal income. 45 In the present study there was a significant association between a sensation-seeking attitude and crime, confirming earlier findings. 46

Community Prevention Efforts
As proposed by Moffit, children and young people are a social group regarded as active in crime and juvenile delinquency. 47 The vast majority of young people commit one or a few, usually less serious, offences during adolescence, but there are relatively few young people who continue on the criminal path later in life. 48 As stated in the introduction, the municipal Social Welfare Boards in Sweden have the ultimate responsibility (beside the youth's guardians) for all children and adolescents. That is, social services have a legally mandated overall responsibility to ensure that all children and youths do have the best living conditions. Hence, social services have a special role in the Swedish welfare system, particularly concerning the welfare of children and adolescents. It might be fair to say that this special role is an effect of the "culture of equality" and "universalism" which has historically shaped the social welfare system in Sweden. 49  system through taxation. 50 The political agenda to reach an equal and inclusive society has historically implied that authorities such as social services use selective and indicated prevention measures to reach groups or individuals in need. This responsibility also includes community efforts to prevent children and adolescents from developing criminal behavior. Preventive measures are, of course, administrated on a voluntary basis. Mandatory measures, such as placement in foster care is only relevant if the behavioral problems, i.e. the youths' own criminal actions or drug use is of a more repeated or serious nature. The most common preventive measure on a selective basis in Swedish municipalities is that social services staff conducts outreach activities in young people's living environments. The purpose is to give children and adolescents, if necessary, access to counseling, support, and protection. In selective, as well as indicated, efforts to target youth criminality, social services have close contact and cooperation with the legal guardians and other authorities, e.g. schools and police. Regarding individually adapted interventions for youth with normbreaking behavior or criminality, a variety of methods are in use. Several include the young person's family, i.e. different types of structured family treatment in outpatient care. 51 The methods are often based on a socio-ecological perspective, e.g. Multisystem Therapy [MST] and Functional Family Therapy [FFT]; a recently updated meta-analysis shows inconsistent results for the former. 52 In a systematic review of 361 studies on crime preventive methods for youth, Mark W. Lipsey reported that therapeutic methods have shown some effect on relapse in criminality. 53 However, in comparison to disciplinary methods such as deterrence, the effects for both groups of methods are quite lowi.e. about 40-50 percent relapse in criminality. 54 Even if the relapse rate indicates that "only" about half of the participants refrain from further criminality after treatment, it is fair to say that it is far better to intervene than not. Positive results for indicated preventive measures such as foster care or residential care have been shown by Bo Vinnerljung and Marie Sallnäs, who conducted a national registry study and analyzed long-term outcomes. 55 About seven hundred Swedish children thirteen to sixteen years old, and placed in foster care or residential care in the early 1990s, were included. At follow-up, conducted at the age of twenty-five, it was shown that the placement itself did not have the desired effect. In particular, youth with behavioral problems showed a higher rate of serious involvement in crime and hospitalizations for mental health problems later in life than those placed in foster care or residential care for other reasons (i.e. abuse by others, not the youth's own actions). 56 In Sweden, at least since the 1980s, incipient youth criminality is addressed with selective or indicated welfare measures and psychosocial care or treatment within the social services. The legislature's view has been, and still is, that youth with behavioral problems, such as criminality, should as far as possible be exempted from criminal sanctions. This position follows the overall "stay out of prison" policy for young offenders. This is also a Criminal Behavior and Mental Health Problems among Adolescents reason why the actions of social services towards this group are primarily legally guided by the principle of voluntariness, and that mandatory measures, even concerning youth at risk, are strictly regulated.
During the latest decade, Sweden has experienced a high rise in gangrelated criminality, foremost in the country's larger cities. This circumstance has led to a slight shift in Swedish criminal policy in another direction, where individuals eighteen and over are more often sentenced to prison than before, thus altering the longstanding "stay out of prison" policy for all juveniles fifteen to twenty years of age. The responsibility of social services for measures to prevent criminality and rehabilitate young offenders are, however, still in place, since the legislature's view hasn't changedonly the judicial tendencies. Young people who commit offenses and are socially disadvantaged are still seen as needing help and support rather than punitive sanctions.

V. Conclusion
We find that the results of our study are of significance for the still ongoing Swedish public and political debate regarding recruitment of children and adolescents into criminal gangs. However, as we see it, the Swedish debate is often one-sided. The debate focuses foremost on the most violent juvenile criminals and the need for increased penal values, but not to the same extent on the overall reasons for youth criminality.
In this study we have firstly shown that mental health problems often are related to youth criminality. Those problems may often be due to factors in the social context. That is, a context where the municipal social services have the primary responsibility for preventive measures. However, some mental health problems may develop into psychiatric problems. This circumstance calls for close collaboration between social services and child and adolescent psychiatry units to detect and help youth in need. Secondly, having a sensation seeking personality and spending time with criminal peers are factors also associated with youth criminality.
These findings confirm previous research that juvenile delinquency can be seen as a group phenomenon that is more easily triggered if one or more persons in a group have a sensation-seeking personality or low control of their impulsivity. 57 Thirdly, better socio-economics and school performance function as protective factors in relation to offending. Allowances for one's own use also function as a protective factor, as does parental education. It should be noted that in this study the protectiveness of school performance, allowances for own use, and parental education differed between genders. These differences have not been reported in previous research. However, the overall interpretation is that the three factors function protectively to hinder youths from committing crimes, if the first two factors are addressed constructively.
In order to balance the one-sided Swedish debate on youth criminality, i.e. the recruitment by criminal gangs, these findings should be noticed. In order to significantly reduce new recruitment by criminal H. Källmen et al.
gangs, as well as to tackle less serious offences from youth at risk, the shown risk and protective factors should be considered by policy makers and legislators both in Sweden and internationally.

Strengths and Limitations
The present study has both strengths and weaknesses, one strength being the large number of respondents from a large proportion of schools in Stockholm, including schools with both a practical vocational orientation and a more theoretical orientation. The most significant limitation is that all variables are self-reported, but this limitation is counterbalanced by the fact that the scales that measure crime and sensation-seeking or anti-social attitudes have good psychometric quality. Another weakness is that the study is based on a cross-sectional design, which means that the results cannot be reliably interpreted in terms of cause and effect. 3 Pratt, "Scandinavian Exceptionalism Part I and II." In Part I Pratt gives attention to the considerably neglected subject of low-imprisonment societies, based on research undertaken in Finland, Norway, and Sweden. By the term "exceptionalism," Pratt firstly refers to the levels of imprisonment in these three countries. In Part II Pratt examines the current prospects for Scandinavian exceptionalism. All three countries have experienced, to a degree, declines in earlier levels of social solidarity, security, and homogeneity which jeopardize the future of the countries' low level of imprisonment.

Notes
4 See Esping-Andersen, Three Worlds of Welfare Capitalism. In short, Esping-Andersen made a still much cited distinction between three types of welfare state regimes: Liberal, Conservative and Social Democratic. 9 See Nilsson, "Confined to the Margins." In short, the objective of this dissertation is to study the living conditions of 141 prison inmates in Sweden from a resource perspective. Findings are organized in three main areas: childhood, living conditions, and recidivism. In comparison to the general population in Sweden inmates can be described as marginalized and/or socially excluded and differ from the general population inter alia in their experience of worse childhood and living conditions. (An abstract in English of the dissertation is available.)