The Development of Delinquency
Research over the past few decades on normal child development and on development of delinquent behavior has shown that individual, social, and community conditions as well as their interactions influence behavior. There is general agreement that behavior, including antisocial and delinquent behavior, is the result of a complex interplay of individual biological and genetic factors and environmental factors, starting during fetal development and continuing throughout life (Bock and Goode, 1996). Clearly, genes affect biological development, but there is no biological development without environmental input. Thus, both biology and environment influence behavior.
Many children reach adulthood without involvement in serious delinquent behavior, even in the face of multiple risks. Although risk factors may help identify which children are most in need of preventive interventions, they cannot identify which particular children will become serious or chronic offenders. It has long been known that most adult criminals were involved in delinquent behavior as children and adolescents; most delinquent children and adolescents, however, do not grow up to be adult criminals (Robins, 1978). Similarly, most serious, chronically delinquent children and adolescents experience a number of risk factors at various levels, but most children and adolescents with risk factors do not become serious, chronic delinquents. Furthermore, any individual factor contributes only a small part to the increase in risk. It is, however, widely recognized that the more risk factors a child or adolescent experiences, the higher their risk for delinquent behavior.
A difficulty with the literature on risk factors is the diversity of the outcome behaviors studied. Some studies focus on behavior that meets diagnostic criteria for conduct disorder or other antisocial behavior disorders; others look at aggressive behavior, or lying, or shoplifting; still others rely on juvenile court referral or arrest as the outcome of interest. Furthermore, different risk factors and different outcomes may be more salient at some stages of child and adolescent development than at others.
Much of the literature that has examined risk factors for delinquency is based on longitudinal studies, primarily of white males. Some of the samples were specifically chosen from high-risk environments. Care must be taken in generalizing this literature to girls and minorities and to general populations. Nevertheless, over the past 20 years, much has been learned about risks for antisocial and delinquent behavior.
This chapter is not meant to be a comprehensive overview of all the literature on risk factors. Rather it focuses on factors that are most relevant to prevention efforts. (For reviews of risk factor literature, see, for example, Hawkins et al., 1998; Lipsey and Derzon, 1998; Rutter et al., 1998.) The chapter discusses risk factors for offending, beginning with risks at the individual level, including biological, psychological, behavioral, and cognitive factors. Social-level risk factors are discussed next; these include family and peer relationships. Finally, community-level risk factors, including school and neighborhood attributes, are examined. Although individual, social, and community-level factors interact, each level is discussed separately for clarity.
INDIVIDUAL-LEVEL RISK FACTORS
A large number of individual factors and characteristics has been associated with the development of juvenile delinquency. These individual factors include age, gender, complications during pregnancy and delivery, impulsivity, aggressiveness, and substance use. Some factors operate before birth (prenatal) or close to, during, and shortly after birth (perinatal); some can be identified in early childhood; and other factors may not be evident until late childhood or during adolescence. To fully appreciate the development of these individual characteristics and their relations to delinquency, one needs to study the development of the individual in interaction with the environment. In order to simplify presentation of the research, however, this section deals only with individual factors.
Studies of criminal activity by age consistently find that rates of offending begin to rise in preadolescence or early adolescence, reach a peak in
late adolescence, and fall through young adulthood (see, e.g., Farrington, 1986a; National Research Council, 1986). Some lawbreaking experience at some time during adolescence is nearly universal in American children, although much of this behavior is reasonably mild and temporary. Although the exact age of onset, peak, and age of desistance varies by offense, the general pattern has been remarkably consistent over time, in different countries, and for official and self-reported data. For example, Farrington (1983, 1986a), in a longitudinal study of a sample of boys in London (the Cambridge Longitudinal Study), found an eightfold increase in the number of different boys convicted of delinquent behavior from age 10 to age 17, followed by a decrease to a quarter of the maximum level by age 24. The number of self-reported offenses in the same sample also peaked between ages 15 and 18, then dropped sharply by age 24. In a longitudinal study of boys in inner-city Pittsburgh (just over half the sample was black and just under half was white), the percentage of boys who self-reported serious delinquent behavior rose from 5 percent at age 6 to about 18 percent for whites and 27 percent for blacks at age 16 (Loeber et al., 1998). A longitudinal study of a representative sample from high-risk neighborhoods in Denver also found a growth in the self-reported prevalence of serious violence from age 10 through late adolescence (Kelley et al., 1997). Females in the Denver sample exhibited a peak in serious violence in midadolescence, but prevalence continued to increase through age 19 for the boys. The study is continuing to follow these boys to see if their prevalence drops in early adulthood. Laub et al. (1998), using the Gluecks' data on 500 juvenile offenders from the 1940s, found that only 25 percent of them were still offending by age 32.
Much research has concentrated on the onset of delinquency, examining risk factors for onset, and differences between those who begin offending early (prior to adolescence) versus those who begin offending in midadolescence. There have been suggestions that early-onset delinquents are more likely than later-onset delinquents to be more serious and persistent offenders (e.g., Moffitt, 1993). There is evidence, however, that predictors associated with onset do not predict persistence particularly well (Farrington and Hawkins, 1991). There are also important problems with the choice of statistical models to create categories of developmental trajectories (Nagin and Tremblay, 1999).
Research by Nagin and Tremblay (1999) found no evidence of late-onset physical aggression. Physical aggression was highest at age 6 (the earliest age for which data were collected for this study) and declined into adolescence. The available data on very young children indicates that frequency of physical aggression reaches a peak around age 2 and then slowly declines up to adolescence (Restoin et al., 1985; Tremblay et al., 1996a).
Those who persist in offending into adulthood may differ from those who desist in a number of ways, including attachment to school, military service (Elder, 1986; Sampson and Laub, 1996), sex, age of onset of offending, incarceration, and adult social bonds (e.g., marriage, quality of marriage, job stability) (Farrington and West, 1995; Quinton et al., 1993; Quinton and Rutter, 1988; Sampson and Laub, 1990). Sampson and Laub (1993) found that marital attachment and job stability significantly reduced deviant behavior in adulthood. Farrington and West (1995) found that offenders and nonoffenders were equally likely to get married, but those who got married and lived with their spouse decreased their offending more than those who remained single or who did not live with their spouse. They also found that offending increased after separation from a spouse. Similarly, Horney et al. (1995) found that married male offenders decreased their offending when living with their spouses and resumed it when not living with them. Within marriages, only good marriages predicted reduction in crime, and these had an increasing effect over time (Laub et al., 1998). Warr (1998) also found that offending decreased after marriage but attributed the decrease to a reduction in the time spent with peers and a reduction in the number of deviant peers following marriage rather than to increased attachment to conventional society through marriage.
Laub et al. (1998) found no difference between persisters and desisters in most family characteristics during childhood (e.g., poverty, parental alcohol abuse or crime, discipline, supervision) or in most individual differences in childhood (e.g., aggression, tantrums, difficult child, verbal IQ). Brannigan (1997) points out that crime is highest when males have the fewest resources, and it lasts longest in those with the fewest investments in society (job, wife, children). Crime is not an effective strategy for getting resources. There is evidence that chronic offenders gain fewer resources than nonoffenders, after the adolescent period (Moffitt, 1993).
The evidence for desistance in girls is not clear. One review of the literature suggests that 25 to 50 percent of antisocial girls commit crimes as adults (Pajer, 1998). There is also some evidence that women are less likely to be recidivists, and that they end their criminal careers earlier than men (Kelley et al., 1997). However, the sexes appear to become more similar with time in rates of all but violent crimes. There is a suggestion that women who persist in crime past adolescence may be more disturbed than men who persist (Jordan et al., 1996; Pajer, 1998).
Prenatal and Perinatal Factors
Several studies have found an association between prenatal and perinatal complications and later delinquent or criminal behavior (Kandel et
al., 1989; Kandel and Mednick, 1991; Raine et al., 1994). Prenatal and perinatal risk factors represent a host of latent and manifest conditions that influence subsequent development.
Many studies use the terms “prenatal or perinatal complications” to describe what is a very heterogeneous set of latent and clinical conditions. Under the heading of prenatal factors, one finds a broad variety of conditions that occurs before birth through the seventh month of gestation (Kopp and Krakow, 1983). Similarly, perinatal factors include conditions as varied as apnea of prematurity (poor breathing) to severe respiratory distress syndrome. The former condition is relatively benign, while the latter is often life-threatening. Although they are risk factors, low birthweight and premature birth do not necessarily presage problems in development.
Prenatal and perinatal risk factors may compromise the nervous system, creating vulnerabilities in the child that can lead to abnormal behavior. Children with prenatal and perinatal complications who live in impoverished, deviant, or abusive environments face added difficulties. According to three major large-scale, long-term studies: (1) developmental risks have additive negative effects on child outcomes, (2) most infants with perinatal complications develop into normally functioning children, and (3) children with long-term negative outcomes who suffered perinatal complications more often than not came from socially disadvantaged backgrounds (Brennan and Mednick, 1997; Broman et al., 1975; Drillien et al., 1980; Werner et al., 1971).
Mednick and colleagues (Brennan and Mednick, 1997; Kandel and Mednick, 1991; Raine et al., 1994) have conducted several investigations in an attempt to elucidate the relationship between criminal behavior and perinatal risk. These and other studies have been unable to identify specific mechanisms to account for the fact that the number of prenatal and perinatal abnormalities tend to correlate with the probability that a child will become a criminal. In addition to the lack of specificity regarding the predictors and the mechanisms of risk, similar measures predict learning disabilities, mental retardation, minimal brain dysfunction, and others (Towbin, 1978). An association between perinatal risk factors and violent offending is particularly strong among offenders whose parents are mentally ill or very poor (Raine et al., 1994, 1997).
Most measures indicate that males are more likely to commit crimes. They are also more vulnerable to prenatal and perinatal stress, as is shown through studies of negative outcomes, including death (Davis and Emory, 1995; Emory et al., 1996).
Hyperactivity, attention problems, and impulsiveness in children have been found to be associated with delinquency. These behaviors can be assessed very early in life and are associated with certain prenatal and perinatal histories (DiPietro et al., 1996; Emory and Noonan, 1984; Lester
et al., 1976; Sameroff and Chandler, 1975). For example, exposure to environmental toxins, such as prenatal lead exposure at very low levels, tends to adversely affect neonatal motor and attentional performance (Emory et al., 1999). Hyperactivity and aggression are associated with prenatal alcohol exposure (Brown et al., 1991; Institute of Medicine, 1996). Prenatal exposure to alcohol, cocaine, heroin, and nicotine appear to have similar effects. Each tends to be associated with hyperactivity, attention deficit, and impulsiveness (Karr-Morse and Wiley, 1997).
Individual Capabilities, Competencies, and Characteristics
In recent investigations, observable behaviors, such as duration of attention to a toy and compliance with mother's instructions not to touch an object, that are particularly relevant to later misbehavior are observable in the first year of life (Kochanska et al., 1998). However, the ability to predict behavior at later ages (in adolescence and adulthood) from such traits early in life is not yet known. Aggressive behavior is nevertheless one of the more stable dimensions, and significant stability may be seen from toddlerhood to adulthood (Tremblay, 2000).
The social behaviors that developmentalists study during childhood can be divided into two broad categories: prosocial and antisocial. Prosocial behaviors include helping, sharing, and cooperation, while antisocial behaviors include different forms of oppositional and aggressive behavior. The development of empathy, guilt feelings, social cognition, and moral reasoning are generally considered important emotional and cognitive correlates of social development.
Impulsivity and hyperactivity have both been associated with later antisocial behavior (Rutter et al., 1998). The social behavior characteristics that best predict delinquent behavior, however, are physical aggression and oppositionality (Lahey et al., 1999; Nagin and Tremblay, 1999). Most children start manifesting these behaviors between the end of the first and second years. The peak level in frequency of physical aggression is generally reached between 24 and 36 months, an age at which the consequences of the aggression are generally relatively minor (Goodenough, 1931; Sand, 1966; Tremblay et al., 1996a, 1999a). By entry into kindergarten, the majority of children have learned to use other means than physical aggression to get what they want and to solve conflicts. Those who have not learned, who are oppositional and show few prosocial behaviors toward peers, are at high risk of being rejected by their peers, of failing in school, and eventually of getting involved in serious delinquency (Farrington and Wikstrom, 1994; Huesmann et al., 1984; Miller and Eisenberg, 1988; Nagin and Tremblay, 1999; Tremblay et al., 1992a, 1994; White et al., 1990).
The differentiation of emotions and emotional regulation occurs during the 2-year period, from 12 months to 36 months, when the frequency of physical aggression increases sharply and then decreases almost as sharply (Tremblay, 2000; Tremblay et al., 1996a, 1999a). A number of longitudinal studies have shown that children who are behaviorally inhibited (shy, anxious) are less at risk of juvenile delinquency, while children who tend to be fearless, those who are impulsive, and those who have difficulty delaying gratification are more at risk of delinquent behavior (Blumstein et al., 1984; Ensminger et al., 1983; Kerr et al., 1997; Mischel et al., 1989; Tremblay et al., 1994).
A large number of studies report that delinquents have a lower verbal IQ compared with nondelinquents, as well as lower school achievement (Fergusson and Horwood, 1995; Maguin and Loeber, 1996; Moffitt, 1997). Antisocial youth also tend to show cognitive deficits in the areas of executive functions1 (Moffitt et al., 1994; Seguin et al., 1995), perception of social cues, and problem-solving processing patterns (Dodge et al., 1997; Huesmann, 1988). The association between cognitive deficits and delinquency remains after controlling for social class and race (Moffitt, 1990; Lynam et al., 1993). Few studies, however, have assessed cognitive functioning during the preschool years or followed the children into adolescence to understand the long-term link between early cognitive deficits and juvenile delinquency. The studies that did look at children 's early cognitive development have shown that poor language performance by the second year after birth, poor fine motor skills by the third year, and low IQ by kindergarten were all associated with later antisocial behavior (Kopp and Krakow, 1983; Stattin and Klackenberg-Larsson, 1993; White et al., 1990). Stattin and Klackenberg-Larsson (1993) found that the association between poor early language performance and later criminal behavior remained significant even after controlling for socioeconomic status.
Epidemiological studies have found a correlation between language delay and aggressive behavior (Richman et al., 1982). Language delays may contribute to poor peer relations that, in turn, result in aggression (Campbell, 1990a). The long-term impact of cognitively oriented preschool programs on the reduction of antisocial behavior is a more direct indication that fostering early cognitive development can play an important role in the prevention of juvenile delinquency (Schweinhart et al., 1993; Schweinhart and Weikart, 1997). It is important to note that since poor cognitive abilities and problem behaviors in the preschool years also
Executive functions refer to a variety of independent skills that are necessary for purposeful, goal-directed activity. Executive functions require generating and maintaining appropriate mental representations, monitoring the flow of information, and modifying problem-solving strategies in order to keep behavior directed toward the goal.
lead to poor school performance, they probably explain a large part of the association observed during adolescence between school failure and delinquency (Fergusson and Horwood, 1995; Maguin and Loeber, 1996; Tremblay et al., 1992).
Several mental health disorders of childhood have been found to put children at risk for future delinquent behavior. Conduct disorder is often diagnosed when a child is troublesome and breaking rules or norms but not necessarily doing illegal behavior, especially at younger ages. This behavior may include lying, bullying, cruelty to animals, fighting, and truancy. Most adolescents in U.S. society at some time engage in illegal behaviors, whether some kind of theft, aggression, or status offense. Many adolescents, in the period during which they engage in these behaviors, are likely to meet formal criteria for conduct disorder. Behavior characterized by willful disobedience and defiance is considered a different disorder (oppositional defiant disorder), but often occurs in conjunction with conduct disorder and may precede it.
Several prospective longitudinal studies have found that children with attention and hyperactivity problems, such as attention deficit hyperactivity disorder, show high levels of antisocial and aggressive behavior (Campbell, 1990b; Hechtman et al., 1984; Loney et al., 1982; Sanson et al., 1993; Satterfield et al., 1982). Early hyperactivity and attention problems without concurrent aggression, however, appear not to be related to later aggressive behavior (Loeber, 1988; Magnusson and Bergman, 1990; Nagin and Tremblay, 1999), although a few studies do report such relationships (Gittelman et al., 1985; Mannuzza et al., 1993, 1991).
Another disorder that is often associated with antisocial behavior and conduct disorder is major depressive disorder, particularly in girls (Kovacs, 1996; Offord et al., 1986; Renouf and Harter, 1990). It is hypothesized that depression during adolescence may be “a central pathway through which girls' serious antisocial behavior develops ” (Obeidallah and Earls, 1999:1). In girls, conduct disorder may be a kind of manifestation of the hopelessness, frustration, and low self-esteem that often characterizes major depression.
For juveniles as well as adults, the use of drugs and alcohol is common among offenders. In 1998, about half of juvenile arrestees in the Arrestee Drug Abuse Monitoring Program tested positive for at least one drug. In these same cities,2 about two-thirds of adult arrestees tested
This program collects information on both juvenile and adult arrestees in Birmingham, Alabama; Cleveland, Ohio; Denver, Colorado; Indianapolis, Indiana; Los Angeles, California; Phoenix, Arizona; Portland, Oregon; St. Louis, Missouri; San Antonio, Texas; San Diego, California; San Jose, California; Tuscon, Arizona; and Washington, DC. Data on adults are collected in 35 cities altogether.
positive for at least one drug (National Institute of Justice, 1999). Of course, drug use is a criminal offense on its own, and for juveniles, alcohol use is also a status delinquent offense. A number of studies have consistently found that as the seriousness of offending goes up, so does the seriousness of drug use as measured both by frequency of use and type of drug (see Huizinga and Jakob-Chien, 1998). In the longitudinal studies of causes and correlates of delinquency in Denver, Pittsburgh, and Rochester (see Thornberry et al., 1995), serious offenders had a higher prevalence of drug and alcohol use than did minor offenders or nonoffenders. In addition, about three-quarters of drug users in each sample were also involved in serious delinquency (Huizinga and Jakob-Chien, 1998). Similarly, in the Denver Youth Survey, serious offenders had the highest prevalence and frequency of use of alcohol and marijuana of all youth in the study. Nevertheless, only about one-third of serious delinquents were problem drug users (Huizinga and Jakob-Chien, 1998).
Although there appears to be a relationship between alcohol and drug use and criminal delinquency, not all delinquents use alcohol or drugs, nor do all alcohol and drug users commit delinquent acts (other than the alcohol or drug use itself). Those who are both serious delinquents and serious drug users may be involved in a great deal of crime, however. Johnson et al. (1991) found that the small group (less than 5 percent of a national sample) who were both serious delinquents and serious drug users accounted for over half of all serious crimes. Neverthless, it would be premature to conclude that serious drug use causes serious crime (McCord, 2001).
Whatever characteristics individuals have, resulting personalities and behavior are influenced by the social environments in which they are raised. Characteristics of individuals always develop in social contexts.
Children's and adolescents' interactions and relationships with family and peers influence the development of antisocial behavior and delinquency. Family interactions are most important during early childhood, but they can have long-lasting effects. In early adolescence, relationships with peers take on greater importance. This section will first consider factors within the family that have been found to be associated with the development of delinquency and then consider peer influences on delinquent behavior. Note that issues concerning poverty and race are dealt with under the community factors section of this chapter. Chapter 7 deals specifically with issues concerning race.
This program collects information on both juvenile and adult arrestees in Birmingham, Alabama; Cleveland, Ohio; Denver, Colorado; Indianapolis, Indiana; Los Angeles, California; Phoenix, Arizona; Portland, Oregon; St. Louis, Missouri; San Antonio, Texas; San Diego, California; San Jose, California; Tuscon, Arizona; and Washington, DC. Data on adults are collected in 35 cities altogether.
In assigning responsibility for childrearing to parents, most Western cultures place a heavy charge on families. Such cultures assign parents the task of raising children to follow society's rules for acceptable behavior. It should be no surprise, therefore, when families have difficulties with the task laid on them, that the product often is juvenile delinquency (Kazdin, 1997). Family structure (who lives in a household) and family functioning (how the family members treat one another) are two general categories under which family effects on delinquency have been examined.
Before embarking on a review of the effects of family structure, it is important to raise the question of mechanisms (Rutter et al., 1998). It may not be the family structure itself that increases the risk of delinquency, but rather some other factor that explains why that structure is present. Alternatively, a certain family structure may increase the risk of delinquency, but only as one more stressor in a series; it may be the number rather than specific nature of the stressors that is harmful.
Historically, one aspect of family structure that has received a great deal of attention as a risk factor for delinquency is growing up in a family that has experienced separation or divorce.3 Although many studies have found an association between broken homes and delinquency (Farrington and Loeber, 1999; Rutter and Giller, 1983; Wells and Rankin, 1991; Wilson and Herrnstein, 1985), there is considerable debate about the meaning of the association. For example, longitudinal studies have found an increased level of conduct disorder and behavioral disturbance in children of divorcing parents before the divorce took place (Block et al., 1986; Cherlin et al., 1991). Capaldi and Patterson (1991) showed that disruptive parenting practices and antisocial personality of the parent(s) accounted for apparent effects of divorce and remarriage. Thus, it is likely that the increased risk of delinquency experienced among children of broken homes is related to the family conflict prior to the divorce or separation, rather than to family breakup itself (Rutter et al., 1998). In their longitudinal study of family disruption, Juby and Farrington (2001) found that boys who stayed with their mothers following disruption had delinquency rates that were almost identical to those reared in intact families.
Many discussions of family structure treat single-parent households and divorced families as the same. In this section, the literature on single-parents is reported separately from that on separated and divorced families because there may be considerable differences in the experiences of children born to single parents and those whose parents divorce.
Being born and raised in a single-parent family has also been associated with increased risk of delinquency and antisocial behavior. Research that takes into account the socioeconomic conditions of single-parent households and other risks, including disciplinary styles and problems in supervising and monitoring children, show that these other factors account for the differential outcomes in these families. The important role of socioeconomic conditions is shown by the absence of differences in delinquency between children in single-parent and two-parent homes within homogeneous socioeconomic classes (Austin, 1978). Careful analyses of juvenile court cases in the United States shows that economic conditions rather than family composition influenced children 's delinquency (Chilton and Markle, 1972). Statistical controls for the mothers' age and poverty have been found to remove effects attributed to single-parent families (Crockett et al., 1993). Furthermore, the significance of being born to a single mother has changed dramatically over the past 30 years. In 1970, 10.7 percent of all births in the United States were to unmarried women (U.S. Census Bureau, 1977). By 1997, births to unmarried women accounted for 32.4 percent of U.S. births (U.S. Census Bureau, 1999). As Rutter and colleagues (1998:185) noted about similar statistics in the United Kingdom: “It cannot be assumed that the risks for antisocial behavior (from being born to a single parent) evident in studies of children born several decades ago will apply to the present generation of births. ” Recent work seems to bear out this conclusion. Gorman-Smith and colleagues found no association between single parenthood and delinquency in a poor, urban U.S. community (Gorman-Smith et al., 1999).
Nevertheless, children in single-parent families are more likely to be exposed to other criminogenic influences, such as frequent changes in the resident father figure (Johnson, 1987; Stern et al., 1984). Single parents often find it hard to get assistance (Ensminger et al., 1983; Spicer and Hampe, 1975). If they must work to support themselves and their families, they are likely to have difficulty providing supervision for their children. Poor supervision is associated with the development of delinquency (Dornbusch et al., 1985; Glueck and Glueck, 1950; Hirschi, 1969; Jensen, 1972; Maccoby, 1958; McCord, 1979, 1982). Summarizing their work on race, family structure, and delinquency in white and black families, Matsueda and Heimer (1987:836) noted: “Yet in both racial groups non-intact homes influence delinquency through a similar process—by attenuating parental supervision, which in turn increases delinquent companions, prodelinquent definitions, and, ultimately, delinquent behavior.” It looks as if the effects of living with a single parent vary with the amount of supervision, as well as the emotional and economic resources that the parent is able to bring to the situation.
A number of studies have found that children born to teenage mothers
are more likely to be not only delinquent, but also chronic juvenile offenders (Farrington and Loeber, 1999; Furstenberg et al., 1987; Kolvin et al., 1990; Maynard, 1997; Nagin et al., 1997). An analysis of children born in 1974 and 1975 in Washington state found that being born to a mother under age 18 tripled the risk of being chronic offender. Males born to unmarried mothers under age 18 were 11 times more likely to become chronic juvenile offenders than were males born to married mothers over the age of 20 (Conseur et al., 1997).
What accounts for the increase in risk from having a young mother? Characteristics of women who become teenage parents appear to account for some of the risk. Longitudinal studies in both Britain and the United States have found that girls who exhibit antisocial behavior are at increased risk of teenage motherhood, of having impulsive liaisons with antisocial men, and of having parenting difficulties (Maughan and Lindelow, 1997; Quinton et al., 1993; Quinton and Rutter, 1988). In Grogger's analysis of data from the National Longitudinal Study of youth, both within-family comparisons and multivariate analysis showed that the characteristics and backgrounds of the women who became teenage mothers accounted for a large part of the risk of their offsprings' delinquency (Grogger, 1997), but the age at which the mother gave birth also contributed to the risk. A teenager who becomes pregnant is also more likely than older mothers to be poor, to be on welfare, to have curtailed her education, and to deliver a baby with low birthweight. Separately or together, these correlates of teenage parenthood have been found to increase risk for delinquency (Rutter et al., 1998). Nagin et al. (1997), in an analysis of data from the Cambridge Study in Delinquent Development, found that the risk of criminality was increased for children in large families born to women who began childbearing as a teenager. They concluded that “the onset of early childbearing is not a cause of children's subsequent problem behavior, but rather is a marker for a set of behaviors and social forces that give rise to adverse consequences for the life chances of children” (Nagin et al., 1997:423).
Children raised in families of four or more children have an increased risk of delinquency (Farrington and Loeber, 1999; Rutter and Giller, 1983). It has been suggested that large family size is associated with less adequate discipline and supervision of children, and that it is the parenting difficulties that account for much of the association with delinquency (Farrington and Loeber, 1999). Work by Offord (1982) points to the influence of delinquent siblings rather than to parenting qualities. Rowe and Farrington (1997), in an analysis of a London longitudinal study, found that there was a tendency for antisocial individuals to have large families. The effect of family size on delinquency was reduced when parents' criminality was taken into account.
Even in intact, two-parent families, children may not receive the supervision, training, and advocacy needed to ensure a positive developmental course. A number of studies have found that poor parental management and disciplinary practices are associated with the development of delinquent behavior. Failure to set clear expectations for children 's behavior, inconsistent discipline, excessively severe or aggressive discipline, and poor monitoring and supervision of children predict later delinquency (Capaldi and Patterson, 1996; Farrington, 1989; Hawkins et al., 1995b; McCord, 1979). As Patterson (1976, 1995) indicates through his research, parents who nag or use idle threats are likely to generate coercive systems in which children gain control through misbehaving. Several longitudinal studies investigating the effects of punishment on aggressive behavior have shown that physical punishments are more likely to result in defiance than compliance (McCord, 1997b; Power and Chapieski, 1986; Strassberg et al., 1994). Perhaps the best grounds for believing that family interaction influences delinquency are programs that alter parental management techniques and thereby benefit siblings as well as reduce delinquent behavior by the child whose conduct brought the parents into the program (Arnold et al., 1975; Kazdin, 1997; Klein et al., 1977; Tremblay et al., 1995).
Consistent discipline, supervision, and affection help to create well-socialized adolescents (Austin, 1978; Bender, 1947; Bowlby, 1940; Glueck and Glueck, 1950; Goldfarb, 1945; Hirschi, 1969; Laub and Sampson, 1988; McCord, 1991; Sampson and Laub, 1993). Furthermore, reductions in delinquency between the ages of 15 and 17 years appear to be related to friendly interaction between teenagers and their parents, a situation that seems to promote school attachment and stronger family ties (Liska and Reed, 1985). In contrast, children who have suffered parental neglect have an increased risk of delinquency. Widom (1989) and McCord (1983) both found that children who had been neglected were as likely as those who had been physically abused to commit violent crimes later in life. In their review of many studies investigating relationships between socialization in families and juvenile delinquency, Loeber and Stouthamer-Loeber (1986) concluded that parental neglect had the largest impact.
Child abuse, as well as neglect, has been implicated in the development of delinquent behavior. In three quite different prospective studies from different parts of the country, childhood abuse and neglect have been found to increase a child's risk of delinquency (Maxfield and Widom, 1996; Smith and Thornberry, 1995; Widom, 1989; Zingraff et al., 1993). These studies examined children of different ages, cases of childhood abuse and neglect from different time periods, different definitions of
child abuse and neglect, and both official and self-reports of offending, but came to the same conclusions. The findings are true for girls as well as boys, and for black as well as for white children. In addition, abused and neglected children start offending earlier than children who are not abused or neglected, and they are more likely to become chronic offenders (Maxfield and Widom, 1996). Victims of childhood abuse and neglect are also at higher risk than other children of being arrested for a violent crime as a juvenile (Maxfield and Widom, 1996).
There are problems in carrying out scientific investigations of each of these components as predictors of juvenile delinquency. First, these behaviors are not empirically independent of one another. Parents who do not watch their young children consistently are less likely to prevent destructive or other unwanted behaviors and therefore more likely to punish. Parents who are themselves unclear about what they expect of their children are likely to be inconsistent and to be unclear in communications with their children. Parenting that involves few positive shared parent-child activities will often also involve less monitoring and more punishing. Parents who reject their children or who express hostility toward them are more likely to punish them. Parents who punish are more likely to punish too much (abuse).
Another problem is the lack of specificity of effects of problems in childrearing practices. In general, problems in each of these areas are likely to be associated with problems of a variety of types —performance and behavior in school, with peers, with authorities, and eventually with partners and offspring. There are also some children who appear to elicit punishing behavior from parents, and this may predate such parenting. Therefore, it is necessary to take account of children's behavior as a potential confounder of the relationship between early parenting and later child problems, because harsh parenting may be a response to a particular child's behavior (Tremblay, 1995). It is also possible that unnecessarily harsh punishment is more frequently and intensely used by parents who are themselves more aggressive and antisocial. Children of antisocial parents are at heightened risk for aggressive, antisocial, and delinquent behavior (e.g., McCord, 1991; Serbin et al., 1998).
Where a family lives affects the nature of opportunities that will be available to its members. In some communities, public transportation permits easy travel for those who do not own automobiles. Opportunities for employment and entertainment extend beyond the local boundaries. In other communities, street-corner gatherings open possibilities for illegal activities. Lack of socially acceptable opportunities leads to frustra-
tion and a search for alternative means to success. Community-based statistics show high correlations among joblessness, household disruption, housing density, infant deaths, poverty, and crime (Sampson, 1987, 1992).
Community variations may account for the fact that some varieties of family life have different effects on delinquency in different communities (Larzelere and Patterson, 1990; Simcha-Fagan and Schwartz, 1986). In general, consistent friendly parental guidance seems to protect children from delinquency regardless of neighborhoods. But poor socialization practices seem to be more potent in disrupted neighborhoods (McCord, 2000).
Neighborhoods influence children's behavior by providing examples of the values that people hold, and these examples influence children's perception of what is acceptable behavior. Communities in which criminal activities are common tend to establish criminal behavior as acceptable. Tolerance for gang activities varies by community (Curry and Spergel, 1988; Horowitz, 1987).
In sum, family life influences delinquency in a variety of ways. Children reared by affectionate, consistent parents are unlikely to commit serious crimes either as juveniles or as adults. Children reared by parents who neglect or reject them are likely to be greatly influenced by their community environments. When communities offer opportunities for and examples of criminal behavior, children reared by neglecting or rejecting parents are more likely to become delinquents. And delinquents are likely to become inadequate parents.
A very robust finding in the delinquency literature is that antisocial behavior is strongly related to involvement with deviant peers. One longitudinal study reported that involvement with antisocial peers was the only variable that had a direct effect on subsequent delinquency other than prior delinquency (Elliott et al., 1985). Factors such as peer delinquent behavior, peer approval of deviant behavior, attachment or allegiance to peers, time spent with peers, and peer pressure for deviance have all been associated with adolescent antisocial behavior (Hoge et al., 1994; Thornberry et al., 1994). In other words, the effects of deviant peers on delinquency are heightened if adolescents believe that their peers approve of delinquency, if they are attached to those peers, if they spend much time with them, and if they perceive pressure from those peers to engage in delinquent acts.
There is a dramatic increase during adolescence in the amount of time adolescents spend with their friends, and peers become increasingly
important during this developmental period. Moreover, peers appear to be most important during late adolescence, with their importance peaking at about age 17 and declining thereafter (Warr, 1993). Thus the decline in delinquency after about age 18 parallels the decline in the importance of peers, including those with deviant influences. Consistent with this view, in the longitudinal research of antisocial British youth by West and Farrington (1977), deviant youth reported that withdrawal from delinquent peer affiliations was an important factor in desistance from offending.
Peer influences appear to have a particularly strong relationship to delinquency in the context of family conflict. For example, adolescents ' lack of respect for their parents influenced their antisocial behavior only because it led to increases in antisocial peer affiliations (Simmons et al., 1991). Patterson et al. (1991) showed that association with deviant peers in 6th grade could be predicted from poor parental monitoring and antisocial activity in 4th grade. And 6th grade association with deviant peers, in turn, predicted delinquency in 8th grade. In adolescence, susceptibility to peer influence is inversely related to interaction with parents (Kandel, 1980; Kandel and Andrews, 1987; Steinberg, 1987).
Other research suggests that adolescents usually become involved with delinquent peers before they become delinquent themselves (Elliott, 1994b; Elliott et al., 1985; Simons et al., 1994). In those cases in which an adolescent was delinquent prior to having delinquent friends, the delinquency was exacerbated by association with deviant peers (Elliott, 1994b; Elliott and Menard, 1996; Thornberry et al., 1993).
The influence of peers varies depending on the influence of parents. In general, peer influence is greater among children and adolescents who have little interaction with their parents (Kandel et al., 1978; Steinberg, 1987). Parents seem to have more influence on the use of drugs among working-class than among middle-class families, and among blacks more than whites (Biddle et al., 1980). Parents also appear to be more influential for the initial decision whether to use any drugs than for ongoing decisions about how and when to use them (Kandel and Andrews, 1987). Patterson and his coworkers emphasize both family socialization practices and association with deviant peers as having strong influences on the onset of delinquency. He hypothesized that “the more antisocial the child, the earlier he or she will become a member of a deviant peer group” (Patterson and Yoerger, 1997:152).
Adolescents report an increasing admiration of defiant and antisocial behavior and less admiration of conventional virtues and talents from age 10 to age 18. They also consistently report that their peers are more antisocial and less admiring of conventional virtues than they are. At age 11, boys report peer admiration of antisocial behavior at a level that is equivalent to what peers actually report at age 17 (Cohen and Cohen,
1996). Adolescents may be more influenced by what they think their peers are doing than by what they actually are doing (Radecki and Jaccard, 1995).
Not only may association with delinquent peers influence delinquent behavior, but also committing a crime with others—co-offending—is a common phenomenon among adolescents (Cohen, 1955; Reiss and Farrington, 1991; Reiss, 1988; Sarnecki, 1986). Much of this behavior occurs in relatively unstable pairings or small groups, not in organized gangs (Klein, 1971; Reiss, 1988). The fact that teenagers commit most of their crimes in pairs or groups does not, of course, prove that peers influence delinquency. Such an influence may be inferred, however, from the increase in crime that followed successful organization of gangs in Los Angeles (Klein, 1971). More direct evidence comes from a study by Dishion and his colleagues. Their research points to reinforcement processes as a reason why deviance increases when misbehaving youngsters get together. Delinquent and nondelinquent boys brought a friend to the laboratory. Conversations were videotaped and coded to show positive and neutral responses by the partner. Among the delinquent pairs, misbehavior received approving responses—in contrast with the nondelinquent dyads, who ignored talk about deviance (Dishion et al., 1996). In addition, reinforcement of deviant talk was associated with violent behavior, even after statistically controlling the boys ' histories of antisocial behavior and parental use of harsh, inconsistent, and coercive discipline (Dishion et al., 1997).
The powerful influence of peers has probably not been adequately acknowledged in interventions designed to reduce delinquency and antisocial behavior. Regarding school-based interventions, among the least effective, and at times harmful, are those that aggregate deviant youth without adult supervision, such as in peer counseling and peer mediation (Gottfredson et al., 1998). Furthermore, high-risk youth are particularly likely to support and reinforce one another 's deviant behavior (e.g., in discussions of rule breaking) when they are grouped together for intervention. Dishion and his colleagues have labeled this process “deviancy training,” which was shown to be associated with later increases in substance use, delinquency, and violence (see the review in Dishion et al., 1999). They argued that youth who are reinforced for deviancy through laughter or attention, for example, are more likely to actually engage in deviant behavior. It is evident that intervenors need to give serious attention to the composition of treatment groups, especially in school settings. It may be more fruitful to construct intervention groups so that low- and moderate-risk youth are included with their high-risk counterparts to minimize the possibility of deviancy training and harmful intervention effects.
Studies of gang participants suggest that, compared with offenders who are not gang members, gang offenders tend to be younger when they begin their criminal careers, are more likely to be violent in public places, and are more likely to use guns (Maxson et al., 1985). Several studies have shown that gang membership is associated with high rates of criminal activities (e.g., Battin et al., 1998; Esbensen et al., 1993; Huff, 1998; Thornberry, 1998; Thornberry et al., 1993). These and other studies (e.g., Pfeiffer, 1998) also suggest that gangs facilitate violence. The heightened criminality and violence of gang members seem not to be reducible to selection. That is, gang members do tend to be more active criminals prior to joining a gang than are their nonjoining, even delinquent peers. During periods of gang participation, however, gang members are more criminally active and more frequently violent than they were either before joining or after leaving gangs. Furthermore, some evidence suggests that gang membership had the greatest effects on those who had not previously committed crimes (Zhang et al., 1999). The literature on gang participation, however, does not go much beyond suggesting that there is a process that facilitates antisocial, often violent, behavior. Norms and pressure to conform to deviant values have been suggested as mechanisms. How and why these are effective has received little attention.
School Policies That Affect Juvenile Delinquency
Delinquency is associated with poor school performance, truancy, and leaving school at a young age (Elliott et al., 1978; Elliott and Voss, 1974; Farrington, 1986b; Hagan and McCarthy, 1997; Hawkins et al., 1998; Huizinga and Jakob-Chien, 1998; Kelly, 1971; Maguin and Loeber, 1996; Polk, 1975; Rhodes and Reiss, 1969; Thornberry and Christenson, 1984). To what extent do school policies contribute to these outcomes for high-risk youngsters? This section outlines what is known about the effects of some of the major school policies that have a particular impact on adolescent delinquents and those at risk for delinquency. The topics covered are grade retention, suspension, and expulsion as disciplinary techniques and academic tracking. These are complex topics about which there is a large literature. This section does not attempt to summarize that literature, but rather to highlight issues that appear to affect juvenile criminality.
Grade retention refers to the practice of not promoting students to the next grade level upon completion of the current grade at the end of the
school year. Low academic achievement is the most frequent reason given by teachers who recommend retention for their students (Jimerson et al., 1997).
There is no precise national estimate of the number of youths who experience grade retention, but the practice was widespread in the 1990s. Contrary to the public perception that few students fail a grade (Westbury, 1994), it is estimated that approximately 15 to 19 percent of students experience grade retention.
Despite the intuitive appeal of retention as a mechanism for improving student performance, the retention literature overwhelmingly concludes that it is not as effective as promotion. Smith and Shepard (1987:130) summarize the effects of grade retention as follows:
The consistent conclusion of reviews is that children make progress during the year in which they repeat a grade, but not as much progress as similar children who were promoted. In controlled studies of the effect of nonpromotion on both achievement and personal adjustment, children who repeat a grade are worse off than comparable children who are promoted with their age-mates. Contrary to popular belief, the average negative effect of retention on achievement is even greater than the negative effect on emotional adjustment and self-concept.
Aside from the effectiveness issue, there are other negative consequences of retention. Retention increases the cost of educating a pupil (Smith and Shepard, 1987). According to Smith and Shepard (1987), alternatives to retention, such as tutoring and summer school, are both more effective and less costly. Retention has negative effects on the emotional adjustment of retainees. For example, Yamamoto and Byrnes (1984) reported that next to blindness and the death of a parent, children rated the prospect of retention as the most stressful event they could suffer. Retained students have more negative attitudes about school and develop characteristics of “learned helplessness,” whereby they blame themselves for their failure and show low levels of persistence. There is a consistent relationship between retention and school dropout (Roderick, 1994; Shepard and Smith, 1990). Dropouts are five times more likely to have repeated a grade than nondropouts, and students who repeat two grades have nearly a 100 percent probability of dropping out. Finally, there are issues of fairness and equity, in that males and ethnic minority children are more likely to be retained (Jimerson et al., 1997).
School Suspension and Expulsion
Unlike grade retention, which is a school policy primarily for young children in the early elementary grades who display academic problems,
suspension and expulsion are mainly directed toward older (secondary school) students whose school difficulties manifest themselves as behavioral problems. Both suspension and expulsion are forms of school exclusion, with the latter being presumably reserved for the most serious offenses.
Supporters of suspension argue that, like any other disciplinary action, suspension reduces the likelihood of misbehavior for the period immediately after suspension and that it can serve as a deterrent to other potentially misbehaving students. Opponents of suspension view the consequences of this disciplinary action as far outweighing any potential benefits. Some of the consequences cited include loss of self-respect, increased chances of coming into contact with a delinquent subculture, the vicious cyclical effects of being unable to catch up with schoolwork, and the stigma associated with suspension once the target child returns to school (Williams, 1989). Furthermore, most investigations of school suspensions have found that serious disciplinary problems are quite rarely the cause of suspension (Cottle, 1975; Kaeser, 1979; McFadden et al., 1992). The majority of suspensions in districts with high suspension rates are for behavior that is not threatening or serious.
The probability of being suspended is unequal among students. Urban students have the highest suspension rates, suburban students have the second highest rates, and rural school students have the lowest rates (Wu et al., 1982). Suspension rates also vary according to sex, race, socioeconomic background, and family characteristics. Male students in every kind of school and education level are about three times more likely to be suspended as females. Suspension rates also vary by race. Statistics indicate that minority students are suspended disproportionately compared with their share in the population and their share of misbehavior, and these racial disparities have the greatest impact on black students; their rate of suspension is over twice that of other ethnic groups, including whites, Hispanics, and Asians (Williams, 1989). Furthermore, black students are likely to receive more severe forms of suspension than other students, even for similar behaviors requiring disciplinary action. In one study, for example, white students were more likely to receive in-school suspension than out-of-school suspension, whereas the reverse pattern was true for black students who had violated school rules (McFadden et al., 1992). This inequality in treatment exists even when factors such as poverty, behavior and attitudes, academic performance, parental attention, and school governance are considered. Students at the lower end of the socioeconomic spectrum tend to be more frequently suspended. Many suspended students come from single-parent families in which the parent had less than a 10th grade education.
Suspended students frequently have learning disabilities or inad-
equate academic skills. Wu et al. (1982) noted a positive relationship between the student suspension rate in a school and the average percentage of students of low ability reported by all teachers in a school. Low-ability students are suspended more than expected, given the number of incidents of misbehavior attributed to them. According to Wu et al. (1982), this phenomenon appears to work in either of two ways. If a student's academic performance is below average, the probability of being suspended increases. And if a school places considerable emphasis on the academic ability of its students, the probability of suspension increases.
Although there is not very much recent empirical research on the effects of school suspension, it appears to be especially detrimental to low-achieving students who may misbehave because they are doing poorly in school. Nor does suspension appear to reduce the behavior it is designed to punish. For example, McFadden et al. (1992) reported that the rate of recidivism remained extremely high across all groups of suspended students in their large study of a Florida school district. Less than 1 percent of disciplined youngsters were one-time offenders, 75 percent were cited for one to five subsequent events during the school year, and 25 percent engaged in more than five serious misbehaviors.
There appear to be clear biases in the use of suspension as a disciplinary action, with black students more likely to be the target of this bias. In the McFadden et al. (1992) study, white students were more likely than their black counterparts to be referred for such misbehaviors as truancy, defiance of authority, and fighting. However, it was the black students who were disproportionately more likely to receive the most severe sanctions, including corporal punishment and out-of-school suspension. As these authors state: “Even though black pupils accounted for only 36.7% of the disciplinary referrals, they received 54.1% of the corporal punishment and 43.9% of the school suspensions, but only 23.1% of the internal suspensions. Additionally, 44.6% of all black pupils referred received corporal punishment, compared to only 21.7% of white pupils and 22.7% of Hispanic pupils” (p. 144).
In sum, the literature reveals that school suspension is academically detrimental, does not contribute to a modification of misbehavior, and is disproportionately experienced by black males, among students who misbehave.
In recent years, expulsion has become a part of the debate on school discipline that has accompanied the rising concern about school violence, particularly that related to weapons possession and increasingly defiant, aggressive behavior by students in school. One result of this debate has been what Morrison et al. (1997) refer to as “zero-tolerance ” disciplinary policies. In California, for example, principals and superintendents are legally obligated to recommend expulsion from the school district for any
student who commits certain offenses, such as bringing weapons to school, brandishing a knife at another person, or unlawfully selling illegal drugs (California Department of Education, 1996-Education Code Section 48900). Such a policy may be expected to increase expulsion given that school officials are required to recommend it in these cases.
Characteristics of children who are expelled parallel those of children who are suspended from school. Students who are expelled tend to be in grades 8 through 12 (Bain and MacPherson, 1990; Hayden and Ward, 1996). There is a fairly substantial group of younger schoolchildren expelled from school; most of them come from the higher age range of students in elementary school. Expulsion is, however, primarily a secondary school phenomenon. About 80 to 90 percent of expelled students are boys, urban students are expelled at a higher rate that students from suburban and rural areas, and minority students are more likely to be expelled than white students.
Morrison and D'Incau (1997) specified four factors related to school adjustment that predicted behavior resulting in recommendation for expulsion. The first is academic performance; poor grade point average, particularly in English and math, and low achievement scores appear to be related to behavior that leads to expulsion. The second is attendance; many expelled students were habitual truants. The third is discipline; many students who experienced expulsion had records of previous suspension. The last factor is special education history; approximately 25 percent of expelled students were either currently, in the past, or in the process of being determined as eligible for special education services.
When children are suspended or expelled from school, their risk for delinquency increases. Exclusion from school makes it more difficult for a child to keep up with academic subjects. Furthermore, with extra time out of school, children are likely to have more time without supervision, and therefore be in a situation known to encourage crime. Effects of school suspension seem to extend beyond childhood. Even after accounting for juvenile criminality, in a national sample of male high school graduates, those who had been suspended were more likely to be incarcerated by the age of 30 (Arum and Beattie, 1999).
Academic tracking, also known as “ability grouping” or “streaming,” describes teaching practices whereby students who seem to be similar in ability are grouped together for instruction. The idea is to reduce the range of individual differences in class groups in order to simplify the task of teaching. Informal tracking is common in elementary schools. For example, teachers may divide children into reading groups based on their
reading skills. Some schools divide students into classrooms based on their assumed ability to learn. These groupings typically also set off upper- and middle-class white children from all others. Because of the fluidity of learning, the particular group into which a child is placed reflects the opinions of the person making the placement at least as much as the ability of the child (see Ball et al., 1984).
Unlike retention, which has been employed mostly in elementary school, and suspension and expulsion, which are largely secondary school phenomena, tracking has proliferated at all levels of schooling in American education. According to Slavin (1987), the practice is nearly universal in some form in secondary schools and very common in elementary schools. A good deal of informal evidence shows that when children considered to be slow learners are grouped together, they come to see themselves in an unfavorable light. Such self-denigration contributes to dislike for school, to truancy, and even to delinquency (Berends, 1995; Gold and Mann, 1972; Kaplan and Johnson, 1991).
Reviews of the effects of tracking in secondary school reach four general conclusions, all suggesting that the impact is largely negative for students in low tracks (see Oakes, 1987). Students in the low-track classes show poorer achievement than their nontracked counterparts. Slavin (1990) found no achievement advantage among secondary school students in high- or average-track classes over their peers of comparable ability in nontracked classes. Rosenbaum (1976) studied the effects of tracking on IQ longitudinally and found that test scores of students in low tracks became homogenized, with a lower mean score over time. Furthermore, he found that students in low tracks tend to be less employable and earn lower wages than other high school graduates; they also often suffer diminished self-esteem and lowered aspirations, and they come to hold more negative attitudes about school. These emotional consequences greatly increase the likelihood of dropping out of school and engaging in delinquent behavior (both in and out of school). One of the clearest findings in research on academic tracking in secondary school is that disproportionate numbers of poor and ethnic minority youngsters (particularly black and Hispanic) are placed in low-ability or noncollege prep tracks (Oakes, 1987). Even within the low-ability (e.g., vocational) tracks, minority students are frequently trained for the lowest-level jobs. At the same time, minority youngsters are consistently underrepresented in programs for the talented and gifted. These disparities occur whether placements are based on standardized test scores or on counselor and teacher recommendations. Oakes and other sociologists of education (e.g., Gamoran, 1992; Kilgore, 1991; Rosenbaum, 1976) have argued that academic tracking frequently operates to perpetuate racial inequality and social stratification in American society.
It is quite evident that all of the policies reviewed here are associated with more negative than positive effects on children at risk for delinquency. As policies to deal with low academic achievement or low ability, neither retention nor tracking leads to positive benefits for students who are experiencing academic difficulty and may reinforce ethnic stereotypes among students who do well. As policies to deal with school misbehavior, neither suspension nor expulsion appears to reduce undesired behavior, and both place excluded children at greater risk for delinquency. Furthermore, every policy covered in this overview has been found to impact ethnic minority youngsters disproportionately.
Growing up in an adverse environment increases the likelihood that a young person will become involved in serious criminal activity during adolescence. Existing research points strongly to the relationship between certain kinds of residential neighborhoods and high levels of crime among young people. Research also points to a number of mechanisms that may account for this association between neighborhood and youth crime. While more research is needed to improve understanding of the mechanisms involved, the link between neighborhood environment and serious youth crime is sufficiently clear to indicate a need for close attention to neighborhood factors in the design of prevention and control efforts.
Two different kinds of research point to the importance of social environment in the generation of antisocial behavior and crime. First, research on the characteristics of communities reveals the extremely unequal geographic distribution of criminal activity. Second, research on human development points consistently to the importance of environment in the emergence of antisocial and criminal behavior. While researchers differ on their interpretation of the exact ways in which personal factors and environment interact in the process of human development, most agree on the continuous interaction of person and environment over time as a fundamental characteristic of developmental processes. Although certain persons and families may be strongly at risk for criminal behavior in any environment, living in a neighborhood where there are high levels of poverty and crime increases the risk of involvement in serious crime for all children growing up there.
This section reviews various strands of research on neighborhoods and crime and on the effects of environment on human development for the purpose of evaluating the contributions of neighborhood environment to patterns of youth crime and prospects for its prevention and control.
Neighborhood Concentrations of Serious Youth Crime
Crime and delinquency are very unequally distributed in space. The geographic concentration of crime occurs at various levels of aggregation, in certain cities and counties and also in certain neighborhoods within a given city or county. For example, cities with higher levels of poverty, larger and more densely settled populations, and higher proportions of unmarried men consistently experience higher homicide rates than those that do not share these characteristics (Land et al., 1990). Serious youth crime in recent years has also been concentrated in certain urban areas. At the peak of the recent epidemic of juvenile homicide, a quarter of all apprehended offenders in the entire United States were arrested in just five counties, containing the cities of Los Angeles, Chicago, Houston, Detroit, and New York. In contrast, during that same year, 84 percent of counties in the United States reported no juvenile homicides (Sickmund et al., 1997).
The concentration of serious crime, especially juvenile crime, in certain neighborhoods within a given city is just as pronounced as the concentration in certain cities. A great deal of research over a period of many decades employing a wide range of methods has documented the geographic concentration of high rates of crime in poor, urban neighborhoods. Classic studies established the concentration of arrests (Shaw and McKay, 1942) and youth gang activity (Thrasher, 1927) in poor neighborhoods located in inner cities. This relationship has been confirmed in replication studies over the years (Bordua, 1958; Chilton, 1964; Lander, 1954; Sampson and Groves, 1989).
In addition to this correlation of neighborhood poverty levels and high crime rates at any given time, research has also found that change in neighborhood poverty levels for the worse is associated with increasing rates of crime and delinquency (Schuerman and Kobrin, 1986; Shannon, 1986). The causal relationship between increases in neighborhood poverty and increases in crime can move in either direction. In the earlier stages of the process of neighborhood deterioration, increases in poverty may cause increases in crime, while, in later stages, crime reaches such a level that those who can afford to move out do so, thereby increasing the poverty rate even further.
Other social characteristics of poor urban neighborhoods change over time and between nations. In the early part of the 20th century in the United States, poor urban neighborhoods tended to be quite mixed in ethnicity (e.g., Italian, Irish, Polish, Jewish), reflecting an era of immigration, and were often located in the older, central parts of cities that were expanding rapidly in outward, concentric waves (Shaw and McKay, 1942). Since the 1950s, poor, urban neighborhoods in the United States have
been much more likely to be dominated by a single cultural group. Blacks and Hispanics, in particular, have experienced an extraordinary degree of residential segregation and concentration in the poorest areas of large cities as a result of racial discrimination in labor and housing markets (Massey and Denton, 1993). In their reanalysis of the Chicago data collected by Shaw and McKay (1942), Bursik and Webb (1982) found that after 1950, changing rates of community racial composition provided a better predictor of juvenile delinquency rates than did the ecological variables.
Poverty and residential segregation are not always urban phenomena. American Indians also experience a great degree of residential segregation and poverty, but rather than in cities, they are segregated on poor, rural reservations. Elsewhere in the developed world, residential concentrations of poor people occur on the periphery of large urban areas, rather than in the center. The construction of large public housing estates in England following World War II produced this kind of urban configuration (Bottoms and Wiles, 1986), in contrast to the concentration on inner-city public housing projects in the United States.
Two important qualifications must be noted with respect to the well-documented patterns of local concentrations of crime and delinquency. First, these patterns do not hold true for minor forms of delinquency. Since a large majority of all adolescent males break the law at some point, such factors as neighborhood, race, and social class do not differentiate very well between those who do or do not commit occasional minor offenses (Elliott and Ageton, 1980).
Second, although some areas have particularly high rates of deviance, in no area do all or most children commit seroius crimes (Elliott et al., 1996; Furstenburg et al., 1999). Still, the concentration of serious juvenile crime in a relatively few residential neighborhoods is well documented and a legitimate cause for concern, both to those living in these high-risk neighborhoods and to the wider society.
Neighborhoods as Mediators of Race and Social Class Disparities in Offending
While studies using differing methods and sources of data are not in agreement on the magnitude of differences in rates of involvement in youth crime across racial, ethnic, and social class categories, most research shows that race, poverty, and residential segregation interact to predict delinquency rates. For example, the three most common approaches to measurement—self-report surveys, victimization surveys, and official arrest and conviction statistics—all indicate high rates of serious offending among young black Americans. There is substantial reason to believe
that these disparate offending rates are directly related to the community conditions under which black children grow up. There is no other racial or ethnic group in the United States of comparable size whose members are nearly as likely to grow up in neighborhoods of concentrated urban poverty (Wilson, 1987). Summarizing this situation, Sampson (1987:353-354) wrote: “the worst urban contexts in which whites reside with respect to poverty and family disruption are considerably better off than the mean levels for black communities.” Although there are more poor white than black families in absolute number, poor white families are far less likely to live in areas where most of their neighbors are also poor. Studies that show stronger effects of race than of class on delinquency must be interpreted in light of the additional stresses suffered by poor blacks as a result of residential segregation.
In comprehensive reviews, scholars have found that adding controls for concentrated neighborhood poverty can entirely eliminate neighborhood-level associations between the proportion of blacks and crime rates. Without controls for concentrated poverty, this relationship is quite strong (Sampson, 1997; Short, 1997). Such research strongly indicates that the unique combination of poverty and residential segregation suffered by black Americans is associated with high rates of crime through the mediating pathway of neighborhood effects on families and children.
These deleterious neighborhood effects have been studied mostly with respect to blacks, but, as the United States has experienced renewed immigration, evidence has also begun to point to similar problems among newer groups of immigrants from Asia, Europe, and Latin America. Much of the evidence at this point is contained in ethnographic studies of youthful gang members and drug dealers (Bourgois, 1995; Chin, 1996; Moore, 1978, 1991; Padilla, 1992; Pinderhughes, 1997; Sullivan, 1989; Vigil, 1988; Vigil and Yun, 1990).
Neighborhood-Level Characteristics Associated with High Rates of Crime and Delinquency
Although the relationship between neighborhood poverty and crime is robust over time and space, a number of other social characteristics of neighborhoods are also associated with elevated levels of crime and delinquency. Factors such as concentrations of multifamily and public housing, unemployed and underemployed men, younger people, and single-parent households tend to be linked to higher crime rates (Sampson, 1987; Wilson, 1985). These social characteristics frequently go along with overall high levels of poverty, but they also vary among both poor and nonpoor neighborhoods and help to explain why neighborhoods with similar average income levels can have different rates of crime.
Recent research has also begun to examine the social atmosphere of neighborhoods and has found significant relationships with crime rates. Neighborhoods in which people tell interviewers that they have a greater sense of collective efficacy—the sense that they can solve problems in cooperation with their neighbors if they have to —have lower crime rates, even when controlling for poverty levels and other neighborhood characteristics (Sampson et al., 1997).
The number and type of local institutions have often been thought to have an effect on neighborhood safety, and some research seems to confirm this. High concentrations of barrooms are clearly associated with crime (Roncek and Maier, 1991). One recent study has also found a crime-averting effect of youth recreation facilities when comparing neighborhoods with otherwise very high rates of crime and criminogenic characteristics to one another (Peterson et al., 2000). Since assessing the number, characteristics, and quality of neighborhood institutions is quite difficult, this remains an understudied area of great importance, given its considerable theoretical and practical interest.
One type of pernicious neighborhood institution, the youth gang, has been studied extensively and is clearly associated with, though by no means synonymous with, delinquency and crime. Although it is true that an adolescent's involvement with youth gangs is associated with a greatly increased risk of criminal behavior, that risk also accompanies association with delinquent peer groups more generally. A very high proportion of youth crime, much higher than for adults, is committed by groups of co-offenders (Elliott and Menard, 1996; Miller, 1982). Most of these delinquent peer groups do not fit the popular stereotypes of youth gangs, with the attendant ritual trappings of distinctive group names, costumes, hand signs, and initiation ceremonies (Sullivan, 1983, 1996). The broader category of delinquent peer groups, most of which are not ritualized youth gangs, drives up neighborhood delinquency rates.
Comparative neighborhood studies, examining the presence of delinquent and unsupervised adolescent peer groups, have found that these groups are more likely to be found in poor neighborhoods. The strength of this finding is such that the presence of these groups appears to be one of the major factors connecting neighborhood poverty and delinquency (Elliott and Menard, 1996; Sampson and Groves, 1989).
Although most adolescent co-offending is committed in the context of delinquent peer groups that are not ritualized youth gangs, the emergence of ritualized gangs in a neighborhood appears to be associated with even higher levels of offending than occur when ritualized gangs are not present (Spergel, 1995; Thornberry, 1998). For this reason, the recent spread of youth gangs across the United States is cause for serious concern. In the decade from the mid-1980s through the mid-1990s, youth
gangs emerged in a growing number of cities in the United States, not only in large cities, but also in smaller cities and towns (Klein, 1995; National Youth Gang Center, 1997).
Despite widespread rumors and mass media allegations, this spread of youth gangs does not appear to be the result of systematic outreach, recruitment, and organization from one city to another. The fact that groups calling themselves by similar names, such as Bloods and Crips, have been spreading from city to city may have very little to do with conscious efforts by members of those groups in Los Angeles to build criminal organizations in other cities. Movies and popular music, rather than direct connections between cities, seem to be at least partly responsible for this copying of gang terminology between cities (Decker and Van Winkle, 1996).
Ethnographic Perspectives on Neighborhoods and Development
A second stream of research that examines adolescent development from the perspective of neighborhood environment consists of ethnographic field studies of delinquent individuals and groups growing up in high-crime neighborhoods. These studies range from classic studies conducted in the 1920s and 1930s (Shaw, 1930; Whyte, 1943), through a second wave in the 1960s (Short and Strodtbeck, 1965; Suttles, 1968) and a more recent wave since the late 1980s (Bourgois, 1995; Chin, 1996; Moore, 1978, 1991; Padilla, 1992; Pinderhughes, 1997; Sullivan, 1989; Vigil, 1988; Vigil and Yun, 1990).
Drawing conclusions from these studies about neighborhood effects on child and adolescent development must be approached carefully, because these studies were primarily designed to describe systems of activity and interaction rather than processes of personal development. As a result, there are many limitations on using this body of research for the purpose of examining neighborhood effects on development, chief among them the predominant focus on single, high-crime areas and the focus within those areas on those engaged in delinquent and criminal activity. Because of this double selection on the dependent variables of both area and individual criminal behavior, these studies generally do not allow systematic comparison between high-crime and low-crime areas or between nondelinquent and delinquent youth within areas.
Despite these limitations, the authors of the studies virtually always end up attributing the ongoing nature of delinquent activity in the areas studied to the influences of the local area on development, particularly among males. In other words, studies not designed primarily to examine development appeal to neighborhood-level influences on development in order to explain their findings. These conclusions about neighborhood
influence on development generally emerge from a much closer scrutiny of the social contexts of development made possible by the in-depth approach of case study and qualitative methods (Sullivan, 1998; Yin, 1989).
One exception to the general lack of comparisons across neighborhoods in the ethnographic studies of development is Sullivan's systematic comparison of three groups of criminally active youths in different neighborhoods of New York City. Using this comparative approach, he demonstrated close links between the array of legitimate and illegitimate opportunities in each place and the developmental trajectories of boys who became involved in delinquency and crime. Even though the early stages of involvement were similar in all three areas, youths from the white, working-class area aged out of crime much faster than their black and Hispanic peers living in neighborhoods characterized by racial and ethnic segregation, concentrated poverty, adult joblessness, and single-parent households. The youths from the more disadvantaged areas had less access to employment and more freedom to experiment with illegal activity as a result of lower levels of informal social control in their immediate neighborhoods (Sullivan, 1989).
Neighborhood-Level Concentrations of Developmental Risk Factors
If neighborhood effects are defined as the influence of neighborhood environment on individual development net of personal and family characteristics, then the amount of variation left over to be attributed to neighborhood in a given study can vary a great deal according to the data and methods used. As many researchers note, neighborhood effects may be mediated by personal and family factors (see, e.g., Farrington and Loeber, 1999); however, it is also necessary to examine whether personal and family characteristics are themselves affected by neighborhood environment. To the extent that this is the case, then neighborhoods affect individual development through their effects on such things as the formation of enduring personal characteristics during early childhood and the family environments in which children grow up. From this perspective, efforts such as those described earlier to measure neighborhood effects net of personal and family characteristics may substantially underestimate neighborhood effects as a result of artificially separating personal and family characteristics from those neighborhood environments. Similarly, if the subsets are not separately analyzed, neighborhood effects will be artificially minimized if some, but not all, types of family constellations increase the impact of neighborhood conditions (McCord, 2000).
A number of studies demonstrate neighborhood concentrations of risk factors for impaired physical and mental health and for the development of antisocial behavior patterns. To date, little research has been able
to trace direct pathways from these neighborhood risk factors through child and adolescent development, although some of the larger ongoing studies, such as the Project on Human Development in Chicago Neighborhoods, are collecting the kind of comprehensive data on biological and social aspects of individual development as well as on the characteristics of a large number of ecological areas that could make this kind of analysis possible (Tonry et al., 1991). Nonetheless, existing research does indicate a number of ways in which deleterious conditions for individual development are concentrated at the neighborhood level. Furthermore, the neighborhoods in which they are concentrated are the same ones that have concentrations of serious youth crime. The risks involved begin for individuals in these areas before birth and continue into adulthood. They include child health problems, parental stress, child abuse, and exposure to community violence.
Neighborhoods with high rates of poverty and crime are often also neighborhoods with concentrations of health problems among children. In New York City, for example, there is a high degree of correlation at the neighborhood level of low birthweight and infant mortality with rates of violent death (Wallace and Wallace, 1990). Moffitt (1997) has pointed to a number of conditions prevalent in inner-city neighborhoods that are capable of inflicting neuropsychological damage, including fetal exposure to toxic chemicals, which are disproportionately stored in such areas, and child malnutrition. Thus, even to the extent that some neighborhoods have larger proportions of persons with clinically identifiable physical and psychological problems, these problems may themselves be due to neighborhood conditions. Thus it can be difficult to disentangle individual developmental risk factors from neighborhood risk factors.
Similarly, some parenting practices that contribute to the development of antisocial and criminal behavior are themselves concentrated in certain areas. McLloyd (1990) has reviewed a wide range of studies documenting the high levels of parental stress experienced by low-income black mothers who, as we have already seen, experience an extremely high degree of residential segregation (Massey and Denton, 1993). This parental stress may in turn lead, in some cases, to child abuse, which contributes to subsequent delinquent and criminal behavior (Widom, 1989). Child abuse is also disproportionately concentrated in certain neighborhoods. Korbin and Coulton's studies of the distribution of child maltreatment in Cleveland neighborhoods have shown both higher rates in poorer neighborhoods and a moderating effect of age structure. Using a combination of qualitative and quantitative methods, they showed that neighborhoods with a younger age structure experienced higher rates of child maltreatment, as measured by reported child abuse cases and inter-
views in a subset of the neighborhoods, than other neighborhoods with similar average family income levels (Korbin and Coulton, 1997).
Recent research has begun to demonstrate high levels of exposure to community violence across a wide range of American communities (Singer et al., 1995), but the degree of exposure also varies by community and reaches extraordinary levels in some neighborhoods. Studies in inner-city neighborhoods have found that one-quarter or more of young people have directly witnessed confrontations involving serious, life-threatening acts of violence, while even larger proportions have witnessed attacks with weapons (Bell and Jenkins, 1993; Osofsky et al., 1993; Richters and Martinez, 1993; Selner-O'Hagan et al., 1998). Various outcomes of this kind of exposure to community violence have been identified. The most commonly cited of these include depressive disorders and posttraumatic stress syndrome, but some links have also been found to increases in aggressive and antisocial behavior (Farrell and Bruce, 1997). Experimental research has shown a pathway from exposure to violence to states of mind conducive to and associated with aggressive behavior, particularly a pattern of social cognition characterized as hostile attribution bias, in which people erroneously perceive others' behavior as threatening (Dodge et al., 1990).
Taken together, these studies point to a multitude of physical, psychological, and social stressors concentrated in the same, relatively few, highly disadvantaged neighborhood environments. Besides affecting people individually, these stressors may combine with and amplify one another, as highly stressed individuals encounter each other in crowded streets, apartment buildings, and public facilities, leading to an exponential increase in triggers for violence (Bernard, 1990). Agnew (1999), having demonstrated the effects of general psychological strain on criminal behavior in previous research, has recently reviewed a wide range of studies that point to just such an amplification effect at the community level.
Environmental and Situational Influences
Other aspects of the environment that have been examined as factors that may influence the risk of offending include drug markets, availability of guns, and the impact of violence in the media.
The presence of illegal drug markets increases the likelihood for violence at the points where drugs are exchanged for money (Haller, 1989). The rise in violent juvenile crime during the 1980s has been attributed to the increase in drug markets, particularly open-air markets for crack cocaine (Blumstein, 1995; National Research Council, 1993). Blumstein (1995) points out the coincidence in timing of the rise in drug arrests of
nonwhite juveniles, particularly blacks, beginning in 1985, and the rise in juvenile, gun-related homicide rates, particularly among blacks. As mentioned earlier, Blumstein argues that the introduction of open-air crack cocaine markets in about 1985 may explain both trends. The low price of crack brought many low-income people, who could afford to buy only one hit at a time, into the cocaine market. These factors led to an increase in the number of drug transactions and a need for more sellers. Juveniles provided a ready labor force and were recruited into crack markets. Blumstein (1995:30) explains how this led to an increase in handgun carrying by juveniles:
These juveniles, like many other participants in the illicit-drug industry, are likely to carry guns for self-protection, largely because that industry uses guns as an important instrument for dispute resolution. Also, the participants in the industry are likely to be carrying a considerable amount of valuable product—drugs or money derived from selling drugs—and are not likely to be able to call on the police if someone tries to rob them. Thus, they are forced to provide for their own defense; a gun is a natural instrument.
Since the drug markets are pervasive in many inner-city neighborhoods, and the young people recruited into them are fairly tightly networked with other young people in their neighborhoods, it became easy for the guns to be diffused to other teenagers who go to the same school or who walk the same streets. These other young people are also likely to arm themselves, primarily for their own protection, but also because possession of a weapon may become a means of status-seeking in the community. This initiates an escalating process: as more guns appear in the community, the incentive for any single individual to arm himself increases.
Other researchers concur that juveniles responded to the increased threat of violence in their neighborhoods by arming themselves or joining gangs for self-protection and adopting a more aggressive interpersonal style (Anderson, 1990, 1994; Fagan and Wilkinson, 1998; Hemenway et al., 1996; Wilkinson and Fagan, 1996). The number of juveniles who report carrying guns has increased. In 1990, approximately 6 percent of teenage boys reported carrying a firearm in the 30 days preceding the survey (Centers for Disease Control and Prevention, 1991). By 1993, 13.7 percent reported carrying guns (Centers for Disease Control and Prevention, 1995). Hemenway et al. (1996) surveyed a sample of 7th and 10th graders in schools in high-risk neighborhoods in a Northeastern and a Midwestern city. Of these, 29 percent of 10th grade males and 23 percent of 7th grade males reported having carried a concealed gun, as did 12 percent of 10th grade females and 8 percent of 7th grade females. The overwhelm-
ing majority gave self-defense or protection as their primary reason for carrying weapons. Moreover, juveniles who reported living in a neighborhood with a lot of shootings or having a family member who had been shot were significantly more likely to carry a gun than other students. Additional student surveys also have found that protection is the most common reason given for carrying a gun (e.g., Centers for Disease Control and Prevention, 1993; Sheley and Wright, 1998).
By studying trends in homicide rates, several researchers have concluded that the increase in juvenile homicides during the late 1980s and early 1990s resulted from the increase in the availability of guns, in particular handguns, rather than from an increase in violent propensities of youth (Blumstein and Cork, 1996; Cook and Laub, 1998; Zimring, 1996). Certainly, assaults in which guns are involved are more likely to turn deadly than when other weapons or just fists are involved. The increase in gun use occurred for all types of youth homicides (e.g., family killings, gang-related killings, brawls and arguments). Furthermore, the rates of nonhandgun homicides remained stable; only handgun-related homicides increased.
Public concern about the role of media in producing misbehavior is as old as concern regarding the socialization of children. Although few believe that the media operate in isolation to influence crime, scientific studies show that children may imitate behavior, whether it is shown in pictures of real people or in cartoons or merely described in stories (Bandura, 1962, 1965, 1986; Maccoby, 1964, 1980). Prosocial as well as aggressive antisocial behavior has been inspired through the use of examples (Anderson, 1998; Eisenberg and Mussen, 1989; Eron and Huesmann, 1986; Huston and Wright, 1998; Staub, 1979). Thus media models can be seen as potentially influencing either risk or protectiveness of environments.
In addition to modeling behavior, exposure to media violence has been shown to increase fear of victimization and to desensitize witnesses to effects of violence (Slaby, 1997; Wilson et al., 1998). Children seem particularly susceptible to such effects, although not all children are equally susceptible. Violent video games, movies, and music lyrics have also been criticized as inciting violence among young people. Cooper and Mackie (1986) found that after playing a violent video game, 4th and 5th graders exhibited more aggression in play than did their classmates who had been randomly assigned to play with a nonviolent video game or to no video game. Anderson and Dill (2000) randomly assigned college students to play either a violent or a nonviolent video game that had been matched for interest, frustration, and difficulty. Students played the same game three times, for a total of 45 minutes, after which they played a competitive game that involved using unpleasant sound blasts against
the rival player. After the second time, measures of the accessibility of aggressive concepts showed a cognitive effect of playing violent video games. After the third time, those who had played the violent video game gave longer blasts of the unpleasant sound, a result mediated by accessibility of aggression as a cognitive factor. The authors concluded that violent video games have adverse behavioral effects and that these occur through increasing the aggressive outlooks of participants.
None of these studies, however, finds direct connections between media exposure to violence and subsequent serious violent behavior. Steinberg (2000:37) summarized the literature on media and juvenile violence by noting: “exposure to violence in the media plays a significant, but very small, role in adolescents' actual involvement in violent activity. The images young people are exposed to may provide the material for violent fantasies and may, under rare circumstances, give young people concrete ideas about how to act out these impulses. But the violent impulses themselves, and the motivation to follow through on them, rarely come from watching violent films or violent television or from listening to violent music . . . . I know of no research that links the sort of serious violence this working group is concerned about with exposure to violent entertainment.”
THE DEVELOPMENT OF DELINQUENCY IN GIRLS
Research on the development of conduct disorder, aggression, and delinquency has often been confined to studies of boys. Many of the individual factors found to be related to delinquency have not been well studied in girls. For example, impulsivity, which has been linked to the development of conduct problems in boys (Caspi et al., 1994; White et al., 1994), has scarcely been studied in girls (Keenan et al., in press).
Behavioral differences between boys and girls have been documented from infancy. Weinberg and Tronick (1997) report that infant girls exhibit better emotional regulation than infant boys, and that infant boys are more likely to show anger than infant girls. This may have implications for the development of conduct problems and delinquency. Although peer-directed aggressive behavior appears to be similar in both girls and boys during toddlerhood (Loeber and Hay, 1997), between the ages of 3 and 6, boys begin to display higher rates of physical aggression than do girls (Coie and Dodge, 1998). Girls tend to use verbal and indirect aggression, such as peer exclusion, ostracism, and character defamation (Bjorkqvist et al., 1992; Crick and Grotpeter, 1995), rather than physical aggression. Research by Pepler and Craig (1995), however, found that girls do use physical aggression against peers, but tend to hide it from adults. Through remote audiovisual recordings of children on a play-
ground, they found the rates of bullying by girls and boys to be the same, although girls were less likely than boys to admit to the behavior in interviews.
Internalizing disorders, such as anxiety and depression, are more frequent in girls and may well overlap with their conduct problems (Loeber and Keenan, 1994; McCord and Ensminger, 1997). Theoreticians have suggested that adolescent females may direct rage and hurt inward as a reaction to abuse and maltreatment. These inward-directed feelings may manifest themselves in conduct problems, such as drug abuse, prostitution, and other self-destructive behaviors (Belknap, 1996).
Whether or not the rate of conduct problems and conduct disorder in girls is lower than that in boys remains to be definitively proven. Girls who do exhibit aggressive behavior or conduct disorder exhibit as much stability in that behavior and are as much at risk for later problems as are boys. Tremblay et al. (1992) found equally high correlations between aggression in early elementary school and later delinquency in boys and girls. Boys and girls with conduct disorder are also equally likely to qualify for later antisocial personality disorder (Zoccolillo et al., 1992).
Delinquency in girls, as well as boys, is often preceded by some form of childhood victimization (Maxfield and Widom, 1996; Smith and Thornberry, 1995; Widom, 1989). Some have speculated that one of the first steps in female delinquency is status offending (truancy, running away from home, being incorrigible), frequently in response to abusive situations in the home (Chesney-Lind and Shelden, 1998). Indeed, Chesney-Lind (1997) has written that status offenses, including running away, may play an important role in female delinquency. In what she refers to as the “criminalization of girls' survival strategies,” Chesney-Lind (1989:11) suggests that young females run away from the violence and abuse in their homes and become vulnerable to further involvement in crime as a means of survival. In one community-based longitudinal study, however, a larger proportion of boys than of girls had left home prior to their sixteenth birthday (McCord and Ensminger, 1997). In a long-term follow-up of a sample of documented cases of childhood abuse and neglect, Kaufman and Widom (1999) reported preliminary results indicating that males and females are equally likely to run away from home, and that childhood sexual abuse was not more often associated with running away than other forms of abuse or neglect. However, the motivation for running away may differ for males and females. For example, females may be running away to escape physical or sexual abuse or neglect in their homes. For boys, running away may be an indirect consequence of childhood victimization or may be part of a larger constellation of antisocial and problem behaviors (Luntz and Widom, 1994).
From the small amount of research that has been done on girls, it appears that they share many risk factors for delinquency with boys. These risk factors include early drug use (Covington, 1998), association with delinquent peers (Acoca and Dedel, 1998), and problems in school (Bergsmann, 1994). McCord and Ensminger (1997) found, however, that, on average, girls were exposed to fewer risk factors (e.g., aggressiveness, frequent spanking, low I.Q., first-grade truancy, early leaving home, and racial discrimination) than were boys.
Delinquent girls report experiencing serious mental health problems, including depression and anxiety, and suicidal thoughts. In a study of delinquent girls conducted by Bergsmann (1994), fully half said that they had considered suicide, and some 64 percent of these had thought about it more than once.
In a survey of mental disorders in juvenile justice facilities, Timmons-Mitchell and colleagues (1997) compared the prevalence of disorders among a sample of males and females and found that the estimated prevalence of mental disorders among females was over three times that among males (84 versus 27 percent). The females in the sample scored significantly higher than males on scales of the Milton Adolescent Clinical Inventory, which measure suicidal tendency, substance abuse proneness, impulsivity, family dysfunction, childhood abuse, and delinquent predisposition. Timmons-Mitchell et al. (1997) concluded from these data that incarcerated female juveniles had significantly more mental health problems and treatment needs than their male counterparts.
Teen motherhood and pregnancy are also concerns among female juvenile offenders. Female delinquents become sexually active at an earlier age than females who are not delinquent (Greene, Peters and Associates, 1998). Sexual activity at an early age sets girls up for a host of problems, including disease and teenage pregnancy, that have far-reaching impacts on their lives and health. Teen mothers face nearly insurmountable challenges that undermine their ability to take adequate care of themselves and their families. Dropping out of school, welfare dependence, and living in poor communities are only a few of the consequences of teen motherhood. And the effects are not limited to one generation. Teen mothers are more likely than women who have children in their early 20s to have children who are incarcerated as adults (Grogger, 1997; Nagin et al., 1997; Robin Hood Foundation, 1996).
Although a large proportion of adolescents gets arrested and an even larger proportion commits illegal acts, only a small proportion commits
serious crimes. Furthermore, most of those who engage in illegal behavior as adolescents do not become adult criminals.
Risk factors at the individual, social, and community level most likely interact in complex ways to promote antisocial and delinquent behavior in juveniles. Although there is some research evidence that different risk factors are more salient at different stages of child and adolescent development, it remains unclear which particular risk factors alone, or in combination, are most important to delinquency. It appears, however, that the more risk factors that are present, the higher the likelihood of delinquency. Particular risk factors considered by the panel are poor parenting practices, school practices that may contribute to school failure, and community-wide settings.
Poor parenting practices are important risk factors for delinquency. Several aspects of parenting have been found to be related to delinquency:
neglect or the absence of supervision throughout childhood and adolescence;
the presence of overt conflict or abuse;
discipline that is inconsistent or inappropriate to the behavior; and
a lack of emotional warmth in the family.
School failure is related to delinquency, and some widely used school practices are associated with school failure in high-risk children. These practices include tracking and grade retention, as well as suspension and expulsion. Minorities are disproportionately affected by these educational and social practices in schools.
Both serious crime and developmental risk factors for children and adolescents are highly concentrated in some communities. These communities are characterized by concentrated poverty. Residents of these communities often do not have access to the level of public resources available in the wider society, including good schools, supervised activities, and health services. Individual-level risk factors are also concentrated in these communities, including health problems, parental stress, and exposure to family and community violence. The combination of concentrated poverty and residential segregation suffered by ethnic minorities in some places contributes to high rates of crime.
Although risk factors can identify groups of adolescents whose probabilities for committing serious crimes are greater than average, they are not capable of identifying the particular individuals who will become criminals.
Delinquency is associated with poor school performance, truancy, and leaving school at a young age. Some pedagogical practices may exacerbate these problems. The available research on grade retention and tracking and the disciplinary practices of suspension and expulsion reveal that such policies have more negative than positive effects. For students already experiencing academic difficulty, tracking and grade retention have been found to further impair their academic performance. Furthermore, tracking does not appear to improve the academic performance of students in high tracks compared with similar students in schools that do not use tracking. Suspension and expulsion deny education in the name of discipline, yet these practices have not been shown to be effective in reducing school misbehavior. Little is known about the effects of these policies on other students in the school. Given the fact that the policies disproportionately affect minorities, such policies may unintentionally reinforce negative stereotypes.
Recommendation: Federal programs should be developed to promote alternatives to grade retention and tracking in schools.
Given that school failure has been found to be a precursor to delinquency, not enough research to date has specifically examined school policies, such as tracking, grade retention, suspension, and expulsion in terms of their effects on delinquent behavior in general. It is important that evaluations of school practices and policies consider their effects on aggressive and antisocial behavior, incuding delinquency. This type of research is particularly salient given the concern over school violence. Research on tracking should examine the effects on children and adolescents in all tracks, not only on those in low tracks.
Recommendation: A thorough review of the effects of school policies and pedagogical practices, such as grade retention, tracking, suspension, and expulsion, should be undertaken. This review should include the effects of such policies on delinquency, as well as the effects on educational attainment and school atmosphere and environment.
Prenatal exposure to alcohol, cocaine, heroin, and nicotine is associated with hyperactivity, attention deficit, and impulsiveness, which are risk factors for later antisocial behavior and delinquency. Biological insults suffered during the prenatal period may have some devastating effects on development. Consequently, preventive efforts during the pre-
natal period, such as preventing fetal exposure to alcohol and drugs, may have great benefits. Reducing alcohol and drug abuse among expectant parents may also improve their ability to parent, thus reducing family-related risk factors for delinquency.
Recommendation: Federal, state, and local governments should act to provide treatment for drug abuse (including alcohol and tobacco use) among pregnant women, particularly, adolescents.
Most longitudinal studies of delinquent behavior have begun after children enter school. Yet earlier development appears to contribute to problems that become apparent during the early school years. Much remains to be known about the extent to which potential problems can be identified at an early age.
Recommendation: Prospective longitudinal studies should be used to increase the understanding of the role of factors in prenatal, perinatal, and early infant development on mechanisms that increase the likelihood of healthy development, as well as the development of antisocial behavior.
Research has shown that the greater the number of risk factors that are present, the higher the likelihood of delinquency. It is not clear, however, whether some risk factors or combinations of risk factors are more important than other risk factors or combinations in the development of delinquency. Furthermore, the timing, severity, and duration of risk factors, in interaction with the age, gender, and the environment in which the individual lives undoubtedly affect the behavioral outcomes. A better understanding of how risk factors interact is important for the development of prevention efforts, especially efforts in communities in which risk factors are concentrated.
Recommendation: Research on risk factors for delinquency needs to focus on effects of interactions among various risk factors. In particular, research on effects of differences in neighborhoods and their interactions with individual and family conditions should be expanded.
The panel recommends the following areas as needing particular research attention to increase understanding of the development of delinquency:
Research on the development of language skills and the impact of delayed or poor language skills on the development of aggressive and antisocial behavior, including delinquency;
Research on children's and adolescents' access to guns, in particular handguns, and whether that access influences attitudes toward or fear of crime;
Research on ways to increase children's and adolescents' protective factors; and
Research on the development of physical aggression regulation in early childhood.
Research on delinquency has traditionally focused on boys. Although boys are more likely to be arrested than girls, the rate of increase in arrest and incarceration has been much larger in recent years for girls than boys, and the seriousness of the crimes committed by girls has increased.
Recommendation: The Department of Justice should develop and fund a systematic research program on female juvenile offending. At a minimum, this program should include:
Research on etiology, life course, and societal consequences of female juvenile offending;
Research on the role of childhood experiences, neighborhoods and communities, and family and individual characteristics that lead young females into crime; and
Research on the role of psychiatric disorders in the etiology of female juvenile crime, as well as its role as a consequence of crime or the justice system's response.