Predictors are arranged in five domains: individual, family, school, peer-related, and community and neighborhood factors. The following malleable predictors of violence are discussed in more detail below.
Individual Medical and Physical Factors
Pregnancy and delivery complications. Prenatal and delivery trauma are somewhat predictive of later violence, although findings vary with the research methods used.
Kandel and Mednick (1991) found that 80 percent of violent offenders scored high in delivery complications, compared with 30 percent of property offenders and 47 percent of nonoffenders. However, other studies have not found an association between pregnancy and delivery complications and violence (Denno, 1990; Farrington, 1997). Mednick and Kandel found in an earlier study (1988) that a stable home environment served as a protective factor against prenatal trauma.
Low resting heart rate. This predictor is thought to indicate a fearless temperament or underarousal, which may predispose an individual to aggression and violence (Raine and Jones, 1987). Research indicates that a low resting pulse rate is a weak predictor of violent crime (Farrington, 1998; Wadsworth, 1976).
The evidence currently does not warrant using either of these predictorspregnancy and delivery complications or low resting heart rateto identify youth at risk for violent behavior. More research is needed on these factors and their possible effects on violence.
Individual Psychological Factors
Internalizing disorders (nervousness/withdrawal, worrying, and anxiety). This category of psychological characteristics has a slight negative correlation with (Mitchell and Rosa, 1979), or is unrelated to, later violence (Farrington, 1989).
Hyperactivity, concentration problems, restlessness, and risk taking. Evidence from studies in this meta-analysis consistently suggests a correlation between these problems and later violent behavior.
In a longitudinal study in Sweden, 15 percent of boys with both restlessness and concentration difficulties at age 13 were arrested for violence by age 26. Boys with restlessness and concentration difficulties were five times more likely to be arrested for violence than boys without these characteristics (Klinteberg et al., 1993).
In another study, Farrington (1989) found that teacher ratings of male children's concentration problems and restlessnessincluding difficulty sitting still, the tendency to fidget, and frequent talkativenesspredicted later violence. Concentration problems also predicted academic difficulties, which predict later violence. Multivariate models are needed to understand the pathways leading to violent behavior.
Aggressiveness. Aggressive behavior measured from ages 6 to 13 consistently predicts later violence among males. Many researchers have noted the continuity in antisocial behavior from early aggression to violent crime (Loeber, 1990, 1996; Loeber and Hay, 1996; Olweus, 1979). A study in Orebro, Sweden, found that two-thirds of boys with high teacher-rated aggression scores at ages 10 and 13 had criminal records for violent offenses by age 26. They were more than six times more likely than boys who were not rated aggressive to be violent offenders (Stattin and Magnusson, 1989).
In a sample of African American boys in the Woodlawn area of Chicago, IL, nearly half of the 6-year-old boys who had been rated aggressive by teachers were arrested for violent crimes by age 33, compared with one-third of their nonaggressive counterparts (McCord and Ensminger, 1995). This relationship also held for males in hyperactive samples (Loney, Kramer, and Milich, 1983).
Research results for females are less consistent. McCord and Ensminger (1995) found similar results for males and females; however, Stattin and Magnusson (1989) did not find a relationship between early female aggression and later violent offenses.
Early initiation of violent behavior. Research has shown that early onset of violence and delinquency is associated with more serious and chronic violence (Farrington, 1991; Piper, 1985; Thornberry, Huizinga, and Loeber, 1995; Tolan and Thomas, 1995). Farrington (1995) found that one-half of boys adjudicated delinquent for a violent offense between age 10 and age 16 were convicted of a violent crime by age 24, compared with only 8 percent of juveniles between age 10 and age 16 not adjudicated delinquent for a violent crime as juveniles.
Involvement in other forms of antisocial behavior. Involvement in antisocial behaviors, including stealing and destruction of property (Mitchell and Rosa, 1979); self-reported delinquency, smoking, and early sexual intercourse (Farrington, 1989); and drug selling (Maguin et al., 1995), is associated with a greater risk of violence among males. Robins (1966) found a similar pattern among male psychiatric patients but did not find similar patterns for females.
Beliefs and attitudes favorable to deviant or antisocial behavior. Dishonesty, antisocial beliefs and attitudes, attitudes favorable to violence, and hostility toward police have been found to predict later violence among males. Relationships between these predictors and violence are less consistent for females (Williams, 1994). Prevention programs that help youth develop positive beliefs and standards so that they can reject violence, cheating, and rule breaking may reduce the risk for violence.
Parental criminality. Baker and Mednick (1984) found that men ages 1823 with criminal fathers were 3.8 times more likely to have committed violent criminal acts than those with noncriminal fathers. Farrington (1989) also found that boys who had a parent arrested before their 10th birthday were 2.2 times more likely to commit violent crimes than those with noncriminal parents.
In contrast, Moffitt (1987) found that adults (ages 2952) with criminal parents were not much more likely to be arrested for a violent offense than those with noncriminal parents. Further research is necessary to understand the contribution of parental criminality to child behavior.
The relationship between parental alcoholism and mental illness and children's violent behavior has been examined. McCord (1979) did not find a link between fathers' alcoholism and criminal conduct and their sons' later violence. In a study of male adoptees, Moffitt (1987) found a small and inconsistent relationship between parental mental illness and violence in children.
Child maltreatment. Studies have examined three forms of child maltreatment: physical abuse, sexual abuse, and neglect. Evidence suggests that children who have been physically abused or neglected are more likely than others to commit violent crimes later in life (Widom, 1989; Zingraff et al., 1993; Smith and Thornberry, 1995).
Poor family management practices. Family management practices such as failure to set clear expectations for children's behavior, poor monitoring and supervision, and severe and inconsistent discipline consistently predict later delinquency and substance abuse (Capaldi and Patterson, 1996; Hawkins, Arthur, and Catalano, 1995). In a sample followed up on after 20 years, the McCords found that parents' poor supervision and aggressive discipline predicted their children's convictions for person crimes well into their forties (McCord, McCord, and Zola, 1959; McCord, 1979).
Wells and Rankin (1988) found that boys with very strict parents reported the most violence. Boys with very permissive parents reported the second highest level of violence. Boys with parents who were neither too strict nor too lax reported the least violence. Also, boys whose parents punished them inconsistently, sometimes punishing and sometimes ignoring the same behavior, were more likely to commit an offense against other persons than boys whose parents punished them more consistently. Parental punitiveness or harshness in discipline also predicted later violence.
Farrington (1989) found that poor childrearing; an authoritarian parenting style; poor parental supervision; harsh parental discipline; a cruel, passive, or neglectful parenting attitude; and parental disagreement about childrearing each predicted later violence. Maguin and colleagues (1995) found that poor family management practices when boys were ages 1416 predicted self-reported violence by age 18, although poor family management practices when boys were age 10 did not predict violence at age 18. An analysis of a subsample of the Seattle Social Development Project data found that proactive family management practices at age 14 reduced the likelihood of self-reported violence at age 16 for African American and Caucasian males and females (Williams, 1994).
Low levels of parental involvement. Strong parental involvement can function as a protective factor against violence. Conversely, a lack of parental interaction and involvement with children may increase children's future risk for violence. Williams (1994) found that parent-child communication and involvement at age 14 predicted less self-reported violent behavior at age 16. This relationship was weaker for females than for males. Similarly, Farrington (1989) found that sons whose fathers did not engage in leisure activities with them more often exhibited violent behavior as teenagers and adults and were more likely to be convicted for a violent offense.
Poor family bonding and conflict. Few studies have looked specifically at the relationship between family bonding and violent behavior. Some research has shown a nonsignificant relationship between poor family bonding and violence (Williams, 1994; Elliott, 1994). Studies investigating this link should distinguish between bonding to prosocial versus antisocial or criminal family members (Foshee and Bauman, 1992).
Exposure to high levels of marital and family conflict also appears to increase the risk of later violence (Farrington, 1989; McCord, 1979; Maguin et al., 1995; Elliott, 1994).
Parental attitudes favorable to substance use and violence. Research indicates that parental attitudes favorable to behaviors such as alcohol use predict use of alcohol and drugs by youth (Peterson et al., 1994), but little research has examined the impact of parental attitudes to violence on children's behavior. One study showed that children who at age 10 had parents who were tolerant of violent behavior were more likely to report violent behavior by age 18 (Maguin et al., 1995).
Residential mobility. Little research has focused on the effect of a family's mobility on youth violence. Maguin and colleagues (1995) found that the number of changes in residence in the past year, assessed when boys were age 16, predicted self-reported violent behavior by 18. Residential mobility assessed when boys were age 14, however, did not significantly predict violence at age 18. This discrepancy may indicate that residential moves have short-term effects on behavior, but more research is needed to understand the relationship.
Parent-child separation. Evidence indicates that disruptions of parent-child relationships predict later violent behavior in children. Parent-child separation before age 10 has been found to predict violence (Farrington, 1989; Wadsworth, 1978). Henry and colleagues (1996) found that having a single-parent family when boys were age 13 predicted their convictions for violence by age 18. An association also has been found between leaving home at an early age and high levels of violence in both men and women (McCord and Ensminger, 1995). However, many other factors that also predict violence can contribute to parent-child separations. Multivariate studies are needed to understand the interactions among these factors.
Various aspects of school-related experiences, such as low educational achievement, low interest in education, dropping out of school, truancy, and poor-quality schools, have been hypothesized to contribute to criminal and violent behavior (Hawkins, Farrington, and Catalano, 1998).
Academic failure. Poor academic achievement has consistently predicted later delinquency (Maguin and Loeber, 1996; Denno, 1990). Academic failure in the elementary grades also increases risk for later violent behavior (Farrington, 1989; Maguin et al., 1995). The relationship between poor academic achievement and later violence has been found to be stronger for females than for males.
Low bonding to school. Research generally supports the hypothesis that bonding to school is a protective factor against crime (Catalano and Hawkins, 1996; Hirschi, 1969). Williams (1994) found that school bonding was a stronger protective factor against violence in African American students and in boys and was less linked to violence in Caucasian students and in girls. Maguin and colleagues (1995) found that low commitment to school and low educational aspirations at age 10 did not predict later violence, but at ages 14 and 16 these factors increased the risk for violence. Other researchers have reported that lack of school bonding was not a significant predictor of serious and violent offending (Elliott, 1994; Mitchell and Rosa, 1979).
Truancy and dropping out of school. Farrington (1989) found that youth with high truancy rates at ages 1214 were more likely to engage in violence as adolescents and adults; leaving school before the age of 15 also predicted later violence. Truancy and dropping out may be indicators of low school bonding, but children also may miss school or leave school early for other reasons (Janosz et al., 1996).
Frequent school transitions. Maguin and colleagues (1995) found that youth who had changed schools often in the past year at ages 14 and 16 were more violent at age 18 than those who had not. Conclusions must be drawn carefully, however, because school transitions can be related to other factors that predict violence.
High delinquency rate school. Farrington (1989) found that boys who at age 11 attended schools with high delinquency rates reported more violent behavior than other youth.
Delinquent siblings. Farrington (1989) found that having delinquent siblings by age 10 predicted later convictions for violence. Maguin and colleagues (1995) found that the association between having delinquent siblings and being convicted for violence was stronger when sibling delinquency occurred closer in time to the violent youth's offense and later in that youth's development, indicating that antisocial siblings have a stronger negative influence during their sibling's adolescence than earlier in the child's development. Williams (1994) found that the influence of delinquent siblings was stronger on girls than on boys.
Delinquent peers. Delinquent peers also may have a greater influence on later violence during an individual's adolescence than they do earlier in development (Moffitt, 1993). Research has shown that adolescents whose peers disapproved of delinquent behavior were less likely to report having committed delinquent acts (Elliott, 1994), including sexual assaults (Ageton, 1983).
Gang membership. Battin and colleagues (1998) showed that being a gang member contributes more to delinquency than does having delinquent peers.
Community and Neighborhood Factors
Community factors, including poverty, low neighborhood attachment and community disorganization, the availability of drugs and firearms, exposure to violence and racial prejudice, laws and norms favorable to violence, and frequent media portrayals of violence, may contribute to crime and violence (Brewer et al., 1995).
Poverty. Being raised in poverty has been found to contribute to a greater likelihood of involvement in crime and violence (Sampson and Lauritsen, 1994). Self-reported felony assault and robbery have been found to be twice as common among youth living in poverty as among middle-class youth (Elliott, Huizinga, and Menard, 1989). Low family income predicted self-reported teen violence and convictions for violent offenses in several studies (Farrington, 1989; Wikström, 1985; Hogh and Wolf, 1983; Henry et al., 1996).
Community disorganization. Maguin and colleagues (1995) examined community disorganization and low neighborhood attachment as predictors of violence. Community disorganization (that is, the presence of crime, drug-selling, gangs, and poor housing) was a better predictor of violence than low attachment to a neighborhood.
Availability of drugs and firearms. In one study, a prevalence of drugs and firearms in the community predicted greater variety in violent behaviors at age 18 (Maguin et al., 1995).
Neighborhood adults involved in crime. Maguin and colleagues (1995) found that children who knew many adult criminals were more likely to engage in violent behavior by age 18. More longitudinal studies investigating the influence of this factor on youth violence are needed.
Exposure to violence and racial prejudice. Exposure to violence in the home and elsewhere increases a child's risk for involvement in violent behavior later in life (Paschall, 1996). McCord and Ensminger (1995) also found that African American study participants who reported having experienced racial discrimination committed more violent acts.
Situational factors are the circumstances that surround a violent event and influence the outcome of that event. These factors may be predictors of violent behavior and may include the presence of a weapon, consumption of alcohol or other drugs by the offender or victim, the behavior of bystanders, the motives of the offender, the relationship of the offender to the victim, and the behavior of the victim (Sampson and Lauritsen, 1994; Farrington and Loeber, 1999). However, the contribution of these factors is difficult to assess because data have not been collected from other situations with similar characteristics in which violence did not occur. Longitudinal studies to investigate these situational triggers are needed.
Multiple Predictors and Strength of Prediction
In the Seattle Social Development Project, Herrenkohl and colleagues (in press) investigated the power of diverse factors seen at ages 10, 14, and 16 to predict violent behavior by the age of 18. More than 17 percent of youth committed a violent act by age 18, and 80 percent of them were expected to do so based on significant predictors seen at age 10. Eighty-four percent were expected to do so based on the significant predictors seen at age 16. The results of the Seattle project are described below for each domainindividual, family, school, peers, and community and neighborhood (Herrenkohl et al., in press).