Violent Victimization As a Risk Factor For

U.S. Department of Justice Office of Justice Programs Office of Juvenile Justice and Delinquency Prevention

J. Robert Flores, Administrator December 2002

Violent Victimization as a Risk Factor for Violent A Message From OJJDP Compared with adults, juveniles are disproportionately affected by high Offending Among Juveniles rates of violence as both offenders and victims. Understanding the relationship between victimization and offending is therefore of critical Jennifer N. Shaffer and R. Barry Ruback importance. As a group, juveniles have high rates of vi- and many of these risk factors suggest Examining data from the National olent victimization and violent offending, a opportunities for intervention. Longitudinal Study of Adolescent pattern suggesting that some juveniles are Health, the authors of this Bulletin The Bulletin includes background informa- both victims and perpetrators of violence. found that victims of violence were tion, a brief theoretical discussion, study To explore that hypothesis, this Bulletin significantly more likely than nonvic- methods and findings, conclusions, policy analyzes the relationships between violent tims to become violent offenders. implications, and suggestions for future victimization and violent offending across They also found that violent victim- research. a 2-year period, using data for 5,003 juve- ization and violent offending share niles who participated in the National Lon- many of the same risk factors, such gitudinal Study of Adolescent Health. The Background as previous violent victimization and Bulletin looks at victimization and offend- offending, drug and alcohol use, and ing experiences in subgroups of juveniles Statistical evidence suggests dispropor- depression. These findings are partic- classified by age, gender, race, and level tionately high rates of violence by and ularly important because they suggest that interventions directed at of physical development. It also identifies against juveniles. This evidence comes preventing victimization could also risk and protective factors for victimiza- both from surveys that ask about behav- reduce offending, and vice versa. tion and offending. Key conclusions and iors and victimization experiences and policy implications include the following: from official records. The analysis presented in this Bul- letin provides evidence that peers and ◆ Violent victimization is indeed a warn- Surveys of self-reported behaviors of adolescents and young adults indicate high adults can and do play important roles ing signal for future violent offending in the lives of juveniles. Juveniles who among juveniles. Protecting juveniles rates of offending among these age groups (Elliott et al., 1983; Lauritsen, Sampson, said that they had support from friends, against violent victimization may, there- parents, teachers, and others were and Laub, 1991). Similarly, surveys of vic- fore, reduce overall levels of juvenile less likely to commit a violent offense. tims’ perceptions of offender characteris- violence. These findings underscore the need tics indicate that the most common age for and value of mentoring, parenting, ◆ Because some groups are at higher risk group for offenders committing rape, rob- than others for violent victimization, and anger management programs bery, and assault is youth ages 18–20, fol- that provide opportunities for juveniles policies and programs aimed at prevent- lowed by juveniles ages 15–17 (Hindelang, to interact with caring adults. By iden- ing victimization may be most effective 1981). Furthermore, Uniform Crime Report tifying youth who are most at risk and if they are focused on these groups. data show that arrest rates for murder, examining the links between victim- ◆ Violent victimization and violent offend- forcible rape, robbery, and aggravated ization and offending, we can improve ing share many of the same risk factors, assault are higher for older teens than for our ability to intervene positively in these juveniles’ lives.

Access OJJDP publications online at ojjdp.ncjrs.org any other age group (Federal Bureau of victimization to the police (Finkelhor and Data Source Investigation, 2000). Ormrod, 1999). The findings reported in this Bulletin When asked about their victimization Research findings are consistent with these are based on statistical analyses of the experiences in the previous year, 18 per- theoretical reasons for expecting that the restricted-access contractual dataset from cent of a large national sample of 8th, 10th, same individuals are often both victims the first two waves of the National Longi- and 12th grade students said they had and offenders. Studies using British Crime tudinal Study of Adolescent Health (known been injured by an attacker who did not Survey data have found a strong positive as the Add Health Study), which is a longi- use a weapon, and 5 percent said they had association between offending and person- tudinal study of a representative national been injured by an attacker with a weapon al victimization among adults (Hough and sample of juveniles in grades 7 through (Johnston, Bachman, and O’Malley, 2001). Mayhew, 1983; Sampson and Lauritsen, 12.1 The study used a clustered sampling Rates of serious violent victimization are 1990). Studies of juveniles in the United design based on a stratified sample of twice as high for juveniles ages 12–17 as States also show that the individuals most 80 high schools and 52 paired middle for adults age 18 or older, and rates of sim- likely to be victims of personal crime are schools.2 Students in these 132 schools ple assault victimization are three times those who report the greatest involvement were asked to complete an in-school ques- higher (Snyder and Sickmund, 1999). in delinquent activities (Lauritsen, Samp- tionnaire. In addition, a subsample, strati- son, and Laub, 1991). In addition, the fied by grade and gender, was selected for greater the variety of delinquent activities, in-home interviews, which included infor- Theoretical Perspective the greater the risk of victimization (e.g., mation about family composition and According to both lifestyle exposure theory Jensen and Brownfield, 1986; Esbensen dynamics, substance use, criminal and and routine activities theory (Hindelang, and Huizinga, 1991; Lauritsen, Sampson, delinquent activities, and violent victim- Gottfredson, and Garofalo, 1978; Cohen and Laub, 1991). ization. The in-home interviews, which and Felson, 1979), individuals’ risk of were conducted in 1995 (year 1) and again criminal victimization depends on their in 1996 (year 2), are the basis for the analy- exposure or proximity to offender popula- Data and Methods ses in this Bulletin. tions, and exposure, in turn, depends on Although earlier studies suggest that crim- The analyses reflect interview data for individuals’ lifestyles and routine activi- inal victimization and criminal offending 5,003 juveniles: 2,402 males and 2,601 ties. Because individuals are most likely are related, the nature of the relationship females; 2,768 non-Hispanic white juve- to interact with those who are similar to is ambiguous. The present study investi- niles and 2,235 minority juveniles;3 1,147 themselves, individuals’ victimization risk gates the nature of the relationship in a juveniles ages 11–14 and 3,856 ages 15–17 is directly proportional to the number of sample of juveniles ages 11–17, addressing at the time of the second interview. The characteristics they share with offenders three issues: analyses exclude respondents who did not (Hindelang, Gottfredson, and Garofalo, ◆ How are violent victimization and vio- have complete data for all of the variables 1978). That is, offenders are more likely lent offending related over time? Does included in the analyses, those whose sec- than nonoffenders to become victims, prior victimization predict subsequent ond interview was conducted less than because their lifestyles frequently bring offending, does prior offending predict 11 months after their first interview,4 and them in contact with other offenders. subsequent victimization, or do they those who were age 18 or older at the Offenders are also more likely than non- both predict each other? In particular, time of the second interview.5 offenders to use alcohol or illegal drugs, is victimization a significant risk factor which lowers their ability to protect them- for subsequent offending? selves and their property, and to live in Analytical Approach neighborhoods characterized by high lev- ◆ What individual-level factors might First, the sample is described in terms of els of population mobility, heterogeneity, explain the relationship between vic- the percentages who reported violent vic- and social disadvantage (e.g., poverty and timization and offending? Do the same timization, violent offending, and both vic- unemployment), which increases their factors predict both violent victimiza- timization and offending, in year 1, year 2, exposure to other offenders (Sampson tion and violent offending? and both years; and the links between victimization and offending within each year and Lauritsen, 1994). ◆ Does drug use affect the relationship are summarized. Next, relationships be- between victimization and offending? Offenders are also likely to be attractive tween violent victimization and violent targets for crime because they can be vic- The study focuses on violence among juve- offending over time are examined—i.e., timized with little chance of legal conse- niles for three reasons. First, from a policy between victimization in year 1 and victim- quences (Sparks, 1982). Offenders are standpoint, it makes sense to concentrate ization in year 2, offending in year 1 and probably less likely than nonoffenders to on the most serious offenses, particularly victimization in year 2, victimization in report victimization to the police because since less is known about the violent vic- year 1 and offending in year 2, and offend- they do not want to draw attention to their timization of juveniles than about the ing in year 1 and offending in year 2—for own illegal behavior (e.g., starting the alter- violent victimization of adults. Second, the sample as a whole and for subgroups cation in question or carrying illegal drugs) because many fewer juveniles engage in based on demographic characteristics and because, if they do file a report, the violence than in property offending and in (age, gender, and race) and level of physi- police probably perceive them as less minor deviant acts, it would be easier to cal development. Finally, through multi- credible than nonoffenders. Offenders’ re- target interventions at this smaller group. variate analyses, the effect of drug use on luctance to report their own victimization Third, the data source for the analyses these relationships is investigated, and might be especially true for violent juve- in this Bulletin included measures of non- risk and protective factors associated with nile offenders, because juveniles in general violent offending but not of nonviolent victimization and offending are explored.6 are less likely than adults to report violent victimization.

2 Measures Findings As shown in table 2, there was a strong The measures of offending and victimiza- link between violent offending and violent tion used in the analyses were dichoto- Incidence victimization within each year. Within year 1, juveniles who offended were 5.3 times mous measures based on juveniles’ yes/ As indicated in table 1, the percentages of more likely than nonoffenders to be vic- no responses to multiple items.7 Two sets offenders and victims were high in years 1 timized (37 percent versus 7 percent), and of measures were used, reflecting the two and 2. Forty percent of juveniles reported those who were victimized were 2.4 times waves of data (i.e., years 1 and 2). The violent offending in year 1, 23 percent in more likely than nonvictims to offend (78 measures of violent offending included year 2, and 17 percent in both years. Nine- percent versus 32 percent). Within year 2, five items reflecting serious physical teen percent reported violent victimization juveniles who offended were 6 times more offenses against other persons: in year 1, 15 percent in year 2, and 9 per- likely than nonoffenders to be victimized cent in both years. Fifteen percent of juve- ◆ Got into a serious physical fight. (42 percent versus 7 percent), and those niles reported both committing and being who were victimized were 4 times more ◆ Hurt someone badly enough to need the victim of a violent crime in year 1, 10 likely than nonvictims to offend (66 percent bandages or care from a doctor or nurse. percent in year 2, and 6 percent in both versus 16 percent). It should be kept in ◆ years. Generally, the percentages of juve- Used or threatened to use a weapon mind that, although the Add Health Study niles reporting offending and victimization to get something from someone. data indicate a close temporal proximity were greater in year 1 than in year 2. This ◆ of offending and victimization, temporal Shot or stabbed someone. decline is likely related to “telescoping,” ordering of events within a year—i.e., ◆ Pulled a knife or gun on someone. or the tendency of survey respondents to whether a particular youth first was vic- report events that occurred outside the The measures of violent victimization timized and then offended, or vice versa— time period about which they were asked included four items reflecting serious cannot be determined from the data. (in this study, prior to year 1).8 physical violence: ◆ Someone pulled a knife or gun on you. ◆ You were shot. Table 1: Incidence of Violent Offending and Violent Victimization ◆ You were cut or stabbed. Percentage of Juveniles Reporting* ◆ You were jumped. Violent Violent Both Offending Juveniles were categorized as offenders if Year Offending Victimization and Victimization they reported committing any of the listed offenses and as nonoffenders if they report- Year 1 40 19 15 ed not committing any of these offenses. Year 2 23 15 10 Juveniles were similarly categorized as Both years 17 9 6 victims or nonvictims, based on whether * Sample size = 5,003. they reported having any of the listed acts committed against them. Juveniles were also categorized on the basis of their reports of using any one Table 2: Relationship Between Violent Offending and Violent Victimization of the following drugs: Within Years: Total Sample

◆ Marijuana. Year 1 ◆ Cocaine. Status in Year 1 Violent Offending (%) Violent Victimization (%) ◆ Inhalants. All (N=5,003) 40 19 ◆ Other drugs, including LSD, PCP, ecstasy, ice (crystal methamphetamine), heroin, Offender 100 37 mushrooms, speed (amphetamines), or Nonoffender 0 7 pills without a doctor’s prescription. Victim 78 100 Juveniles were categorized into one of four Nonvictim 32 0 groups: nonusers (no reported use of any drug at either of the two interviews), de- Year 2 sisters (reported use at first interview but not at the second interview), new users Status in Year 2 Violent Offending (%) Violent Victimization (%) (reported use at the second interview but All (N=5,003) 23 15 not the first), and consistent users (reported use at both interviews). Juveniles were Offender 100 42 similarly categorized on the basis of Nonoffender 0 7 reported alcohol use. Victim 66 100 Nonvictim 16 0

3 By age. Juveniles were divided into two commit a violent offense in year 2 and to be Key Findings groups according to their age in year 2: victims of violence in year 2. For both males 11–14 and 15–17. As shown in table 4, the and females, juveniles who offended in year ◆ Juveniles who were victims of vio- total percentage of juveniles who were vic- 1 were significantly more likely than non- lence in year 1 were significantly tims of violence in year 2 was significantly offenders to be victimized in year 2, and more likely than nonvictims to com- greater for the older group; however, there those who were victimized in year 1 were mit a violent offense in year 2 and was no significant difference between the significantly more likely than nonvictims to to be victims of violence in year 2. two age groups in the percentages who offend in year 2. Again, these findings are ◆ Juveniles who committed a violent committed a violent offense in year 2. For consistent with the pattern for all juveniles. offense in year 1 were significantly both age groups, juveniles who offended in year 1 were significantly more likely than By race. The data were analyzed separate- more likely than nonoffenders to ly for white and minority juveniles (see commit a violent offense in year 2 nonoffenders to be victimized in year 2, and those who were victimized in year 1 endnote 3 for racial groups included in the and to be victims of violence in minority category). As indicated in table 6, year 2. were significantly more likely than nonvictims to offend in year 2. These findings are both the total percentage of juveniles who ◆ In general, these patterns were consistent with the pattern for all juveniles. committed a violent offense in year 2 and true regardless of age, gender, the total percentage who were victims of race, level of physical develop- By gender. As shown in table 5, males were violence in year 2 were greater for the mi- ment, or drug use. significantly more likely than females to nority category. For both minorities and

Because of this limitation, the rest of the Table 3: Relationship Between Violent Offending and Violent Victimization analyses presented in this Bulletin will Across Years: Total Sample focus on the relationships between violent victimization and offending across years. Year 2 Status in Year 1 Violent Offending (%) Violent Victimization (%) Relationships Among Variables All (N=5,003) 23 15 The data in tables 3–7 indicate relationships Offender 44 28 between violent offending and violent vic- Nonoffender 10 6 timization across years. Table 3 shows these relationships for the total sample. Tables Victim 52 47 4–7 show the relationships by age group, Nonvictim 17 8 gender, race, and level of physical development. In the discussion that follows, all references to “significant” differences in respondents’ likelihood of offending or Table 4: Relationship Between Violent Offending and Violent Victimization victimization (i.e., differences between Across Years, by Age Group percentages reported in the tables) refer to differences that are statistically signifi- Year 2 cant at the p < .05 level, based on standard chi-square tests. Readers should use cau- Status in Year 1 Violent Offending (%) Violent Victimization (%) tion when comparing estimates not explic- Ages 11–14 itly discussed in the text; what may appear to be a large difference may not be a sta- All (n=1,147) 23 13 tistically significant difference. Offender 42 23 Total sample. Table 3 shows that, in the Nonoffender 12 6 total sample of 5,003 juveniles, those who Victim 56 37 committed a violent offense or were vic- Nonvictim 18 8 tims of violence in year 1 had a significantly increased likelihood of offending or being Ages 15–17 victimized in year 2. Juveniles who offended in year 1 were 4.4 times more likely than All (n=3,856) 22 16 nonoffenders to offend in year 2 (44 per- Offender 45 30 cent versus 10 percent) and 4.7 times more Nonoffender 9 7 likely to be victimized in year 2 (28 percent versus 6 percent). Juveniles who were vic- Victim 51 50 timized in year 1 were 3 times more likely Nonvictim 16 7 than nonvictims to offend in year 2 (52 per- Note: Because the analysis focuses on predicting outcomes for juveniles in year 2, the age measure cent versus 17 percent) and 6 times more reflects age at year 2. Thus, juveniles in the 11Ð14 age group were ages 10Ð13 in year 1, and juve- likely to be victimized in year 2 (47 percent niles in the 15Ð17 age group were ages 14Ð16 in year 1. versus 8 percent).

4 ◆ For males: Table 5: Relationship Between Violent Offending and Violent Victimization ❖ How much hair is under your arms? Across Years, by Gender ❖ How thick is the hair on your face? Year 2 ❖ How much lower is your voice than when you were in grade school? Status in Year 1 Violent Offending (%) Violent Victimization (%) ❖ How advanced is your physical Males development compared to other All (n=2,402) 30 22 boys your age? ◆ For females: Offender 48 32 Nonoffender 14 9 ❖ How much more developed are your breasts than when you were in Victim 56 51 grade school? Nonvictim 23 11 ❖ How much more curvy is your body Females compared to when you were in grade school? All (n=2,601) 15 19 ❖ How advanced is your physical Offender 38 20 development compared to other Nonoffender 7 4 girls your age? Victim 44 5 As shown in table 7, the total percentage of Nonvictim 12 9 juveniles who committed a violent offense in year 2 and the total percentage who were victims of violence in year 2 were significantly greater for more physically devel- Table 6: Relationship Between Violent Offending and Violent Victimization oped juveniles. For both groups, juveniles Across Years, by Race who offended in year 1 were significantly more likely than nonoffenders to be victim- Year 2 ized in year 2, and those who were victim- Status in Year 1 Violent Offending (%) Violent Victimization (%) ized in year 2 were significantly more likely than nonvictims to offend in year 2. Again, Minority juveniles these findings are consistent with the pattern for all juveniles. All (n=2,235) 25 20 Offender 42 32 Effects of Drug Use Nonoffender 13 12 Drug use and its influence on the relation- Victim 50 53 ship between violent victimization and of- Nonvictim 20 11 fending were also examined. In both years, the percentage of juveniles reporting drug White juveniles use was significantly greater among juve- All (n=2,768) 21 12 niles ages 15–17 than juveniles ages 11–14 but was not significantly different for males Offender 45 25 and females or for minorities and whites. Nonoffender 9 4 To determine the influence of drug use on the relationship between victimization and Victim 54 43 offending, multivariate analysis was per- Nonvictim 16 6 formed within each of the four drug use categories defined on page 3: nonusers (n=3,270); desisters (n=481), new users whites, juveniles who offended in year 1 By level of physical development. The (n=455), and consistent users (n=797). In were significantly more likely than non- data were also analyzed separately by general, drug use did not influence the offenders to be victimized in year 2 and to juveniles’ level of physical development. victimization-offending relationship: vio- offend in year 2, and juveniles who were Juveniles were categorized as “more lent victimization increased the risk of victimized in year 1 were significantly more physically developed” or “less physical- violent offending and violent offending likely than nonvictims to offend in year 2 ly developed,” based on their responses increased the risk of violent victimization, and to be victimized in year 2. Again, these to the following questions at the second regardless of drug use.10 findings are consistent with the pattern for interview:9 all juveniles.

5 Risk and Protective Factors Table 7: Relationship Between Violent Offending and Violent Victimization Multivariate analyses were also used to Across Years, by Level of Physical Development identify risk and protective factors for violent offending and victimization (i.e., fac- Year 2 tors that independently predict offending Status in Year 1 Violent Offending (%) Violent Victimization (%) or victimization in year 2 after statistical controls for other factors are introduced More physically into the model). Tables 8 and 9 present developed juveniles the results of the analyses, and the side- All (n=2,449) 25 17 bar on page 8 discusses the methodology used in the analyses. Offender 50 32 Nonoffender 10 7 Violent offending. The results presented in table 8 indicate that, even after other Victim 57 52 factors related to violent offending were Nonvictim 19 8 controlled statistically, being a victim of a violent crime in year 1 was still a signifi- Less physically cant risk factor for committing a violent developed juveniles offense in year 2. Only violent offending in All (n=2,554) 20 13 year 1 had a greater influence. The analysis also revealed an important protective Offender 38 25 factor against violent offending in year 2: Nonoffender 9 4 juveniles who reported greater support Victim 47 42 from important people in their lives, such Nonvictim 15 7 as friends, parents, and teachers, were less likely to commit a violent offense in year 2. Violent victimization. The results presented in table 9 indicate that, when all other Table 8: Factors Predicting Violent Offending in Year 2 risk factors were controlled statistically, committing a violent offense in year 1 was † ‡ Predictor* Logistic Coefficient Odds Ratio still a significant risk factor for being the Violent offending in year 1 1.39 (.09) 4.01 victim of a violent crime in year 2. Only the effects of violent victimization in year 1, Violent victimization in year 1 0.86 (.11) 2.36 being male, being a consistent drug user, and being a new drug user had a greater Male 0.71 (.11) 2.03 influence. Finally, violent victimization was Consistent drug user 0.62 (.16) 1.86 significantly less likely among white juveniles than among minority juveniles,11 New alcohol user 0.59 (.15) 1.80 among juveniles who resided in two-parent Consistent alcohol user 0.56 (.13) 1.75 households than among those with other family structures, and among juveniles New drug user 0.36 (.17) 1.43 who resided in households with higher More physically developed 0.30 (.10) 1.35 socioeconomic status than among those with lower socioeconomic status. Depression 0.23 (.11) 1.26 Support from significant others –0.26 (.09) 0.77 Conclusions Household socioeconomic status –0.19 (.06) 0.83 The analyses suggest three major conclusions: Note: See sidebar on page 8 for methodological notes. Violent victimization is an important risk * Only significant (p < .05) predictors are reported here. The sidebar on page 8 includes a complete factor for subsequent violent offending. list of the variables analyzed. The percentage of year 1 victims who committed a violent offense in year 2 (52 † The logistic coefficient represents the effect of a given predictor variable (e.g., violent offending in year 1) on the log odds of the outcome (i.e., violent offending in year 2). Positive numbers indicate percent) was significantly higher than risk factors; negative numbers indicate protective factors. Standard errors are in parentheses. the percentage of year 1 nonvictims who committed a violent offense in year 2 (17 ‡ The odds ratio indicates the proportional change in the odds of violent offending in year 2, per percent). The figure on page 7 (which one-unit increase in the predictor variable. The greater the difference from one, the greater the reflects the percentages in table 3) illus- effect of the variable on violent offending. trates this finding.

6 Repeat offending is more common than Table 9: Factors Predicting Violent Victimization in Year 2 repeat victimization. The relationship of violent offending in year 1 to violent of- Predictor* Logistic Coefficient† Odds Ratio‡ fending in year 2 was stronger than the relationship of violent victimization in Violent victimization in year 1 1.74 (.13) 5.70 year 1 to violent victimization in year 2. Male 0.91 (.13) 2.48 Approximately twice as many juveniles committed an offense in both years as Consistent drug user 0.85 (.18) 2.34 were victimized in both years. New drug user 0.77 (.19) 2.16 Violent victimization and violent offend- Violent offending in year 1 0.69 (.12) 1.99 ing share many of the same risk factors. Shared risk factors include previous violent Depression 0.52 (.12) 1.68 victimization and offending, use of drugs or Consistent alcohol user 0.47 (.16) 1.60 alcohol, being male, depression, and having a high level of physical development. Easy access to gun in home 0.40 (.16) 1.49 More physically developed 0.29 (.13) 1.34 Policy Implications Time spent hanging out with friends 0.17 (.06) 1.19 The findings of this analysis have at least four policy implications: White –0.53 (.14) 0.59 Some groups are at higher risk than Two-parent household –0.43 (.13) 0.65 others for violent victimization. The Household socioeconomic status –0.23 (.07) 0.79 percentage of juveniles who were victims of violent crime in this sample was high: Note: See sidebar on page 8 for methodological notes. 26 percent were victimized at least once during the 2-year study period and 9 per- * Only significant (p < .05) predictors are reported here. The sidebar on page 8 includes a complete cent were victimized at least twice. Signifi- list of the variables analyzed. cantly higher rates of violent victimization † The logistic coefficient represents the effect of a given predictor variable (e.g., violent victimization in were found among juveniles with certain year 1) on the log odds of the outcome (i.e., violent victimization in year 2). Positive numbers indicate characteristics—those who used drugs risk factors; negative numbers indicate protective factors. Standard errors are in parentheses. consistently or began to use drugs, those who were depressed, members of racial ‡ The odds ratio indicates the proportional change in the odds of violent victimization in year 2, per one-unit increase in the predictor variable. The greater the difference from one, the greater the effect minority groups, and older juveniles who of the variable on violent victimization. committed violent offenses. These findings suggest that victimization prevention programs may be most effective if they are focused on these groups. Because of the strong association between drug use Percentage of Juveniles Who Offended or Were Victimized in Year 2, and victimization, drug use prevention by Status in Year 1 and treatment programs might be promis- ing strategies for decreasing juveniles’ risk 60 of violent victimization. Violent victimization is a warning signal 50 for future violent victimization. About 40 one-half of the juveniles who reported being victims of violence during year 1 30 also reported being victimized during year 2. These repeat victims might be especial- 20 ly suitable for interventions to prevent future victimization. Other research has 10

ercentage of Juveniles shown that crime victims are more likely P 0 than nonvictims to experience depression, Violent No violent Violent No violent anxiety, and physical health problems offender offenses victimization victimization (Kilpatrick et al., 1985). Studies have also shown that the greater the severity of the Status in Year 1 victimization (e.g., a higher level of violence), the more severe the symptoms Violent offending in year 2 Violent victimization in year 2 (Bard and Sangrey, 1985; Riggs, Rothbaum, and Foa, 1995). The current study found

7 that the higher the level of juveniles’ depression, the greater their likelihood of Methodology for Analyses of Risk and Protective Factors becoming victims of violence. This finding suggests that focusing counseling and The multivariate analyses of risk and protective factors used cross-lag logistic re- other victim services on juvenile victims of gression techniques. For each outcome variable at year 2, the statistical model violent crime—especially repeat victims— included a term controlling for the effect of that variable at year 1. For example, the may be particularly important. model predicting the log odds of being a violent offender in year 2 included a control for status as a violent offender in year 1. For reasons similar to those explained Violent victimization is a warning signal in endnote 10, the results of multivariate analyses presented in tables 8 and 9 are for future violent offending. The finding based on models that excluded all interaction terms for age, gender, and race. that being a victim of a violent crime pre- The multivariate analyses of victimization and offending in year 2 included the fol- dicted violent offending suggests that vic- lowing independent variables: time spent hanging out with friends; drug use; alco- timization is itself a risk factor for offend- hol use; tobacco use; race; depression (a standardized, composite index reflecting ing or is correlated with some factor or juveniles’ mean score on 14 psychosomatic symptoms and 5 emotional symptoms process that is a risk factor. This implica- commonly associated with depression); support from others (a standardized, com- tion, in turn, suggests that protecting juve- posite index of 7 items reflecting juveniles’ perceptions of how much adults, friends, niles against violent victimization may and teachers care about them); age (in years); age squared; easy access to a fire- reduce overall levels of juvenile violence. arm in the home; whether the juvenile lived in a home with two parental figures Because juveniles are probably more like- during both years; and household socioeconomic status (a standardized mean of ly to admit victimization than offending, parental occupational prestige and parental education). Variables that were not interventions focused on victims might be significant predictors are excluded from tables 8 and 9. easier to accomplish than interventions The models predicting victimization in year 2 included controls for juveniles’ prop- focused on offenders. The finding that erty offending and minor deviance and delinquency. This was done to ensure that the effect of violent victimization on offend- the observed effects of violent offending on violent victimization reflected only the ing appears to be stronger within years effect of violence and not the tendency of juveniles who commit violent crimes to than across years (see tables 2 and 3) also commit property crimes and to be involved in other delinquent and deviant suggests that interventions may be most activities. The measure of property offending was a dichotomous variable based successful in preventing future offending on six activities reflecting involvement in property offending. The measure of minor if they are applied relatively soon after deviance and delinquency was also a dichotomous variable and was based on five the victimization. activities reflecting involvement in minor crimes (e.g., disorderly conduct) and status offenses (e.g., running away from home). Many of the risk factors associated with juvenile violence suggest opportunities for intervention. A number of the risk factors presented in tables 8 and 9 involve themselves. Research should also focus should provide more rigorous tests of the behavior of juveniles and people who on the extent to which victimization affects measures of routine activities to explain are important in their lives; as such, these juveniles’ mental and physical health. the relationship between offending and factors are appropriate points for inter- victimization in juveniles. vention. Because the majority of risk fac- The role of delinquent peers should be tors predicted both violent offending and investigated. Because in most crimes the The reciprocal nature of the relationship violent victimization, it may be possible victim and the offender know each other, between victimization and offending for interventions to simultaneously reduce association with delinquent peers may needs clarification. The present study juveniles’ risk of both. explain not only juveniles’ offending found some evidence that the relationships (Hawkins et al., 2000) but also their vic- between victimization and offending in timization. Previous research suggests juveniles may be more simultaneous than Future Research that some of the relationship between the cross-year relationships that were the This study suggests several areas for offending and victimization and between focus of the study. Future research should future research: victimization and offending is explained explore the possibility that violent victim- by the delinquency of juveniles’ peers ization and offending may have stronger The links by which offending and victim- (Lauritsen, Sampson, and Laub, 1991; short-term than long-term influences on ization affect each other should be ex- Fagan, Piper, and Cheng, 1987). However, each other. plored. The findings of this study suggest the effect of peers has not been isolated both that offending increases juveniles’ from the effects of juveniles’ own prior risk of victimization and that victimization offending and victimization. A related find- For Further Information increases juveniles’ risk of offending. Fu- ing of the present study—that time spent To obtain data files from the National Lon- ture research should examine links that in unstructured activity with peers (i.e, gitudinal Study of Adolescent Health, con- might explain these two processes. For “hanging out with friends”) was a signifi- tact Jo Jones, Carolina Population Center, example, being victimized might increase cant risk factor for violent victimization— 123 West Franklin Street, Chapel Hill, NC juveniles’ use of drugs, and drug use might offers some evidence to support victimiza- 27516–3997 (e-mail [email protected]). make them more vulnerable to victimization theories based on routine activities tion because they are less able to protect and lifestyle exposure. Future research

8 Endnotes 7. The response alternatives for offending and offending and victimization, were test- and victimization were “never,” “once,” ed. The number of significant interactions 1. The Add Health Study is being conduct- and “more than once.” For the analyses was smaller than would be expected to ed by the Carolina Population Center at the presented here, these responses were occur by chance, and the few significant University of North Carolina at Chapel Hill recoded to “yes” (once or more than interactions had no clear pattern within or under a grant from the National Institute once) or “no” (never). It is important to across the categories of drug use. More- of Child Health and Human Development. keep in mind that most of the measures over, within the specific distribution cells Approval to use the data was granted by discussed in this Bulletin necessarily rely (e.g., female consistent drug users), the the Carolina Population Center, the Popu- on the willingness and accuracy of the sample sizes were quite small, resulting in lation Research Institute at Pennsylvania juvenile respondents’ self-reports. Analy- less stable estimates and lower statistical State University, and the Pennsylvania ses of the National Crime Victimization power for detecting significant effects. State University Institutional Review Board. Survey (NCVS) indicate that juveniles Thus, the results presented here are based 2. In the Add Health Study, clusters were are less likely than adults to report vio- on models that excluded all interaction sampled with unequal probability. Although lent victimization to the police, although terms. this method reduced the costs of data juveniles and adults are equally likely 11. In the analyses of risk and protective collection, the design complicated the sta- to report theft offenses (Finkelhor and factors for both offending and victimiza- tistical analysis because the observations Ormrod, 1999). In addition, the Add Health tion, it is important to interpret findings were not independent and identically dis- Study interview did not include the spe- regarding juveniles’ race cautiously. The tributed. Correct analysis of the data re- cial rapport-building and screener ques- analyses presented here did not include quires the use of special survey software tions currently used in the NCVS to cap- controls for juveniles’ neighborhoods, and packages capable of handling observations ture violent crimes against women. Finally, it is possible that the observed effects for that are not independent and not identical- the Add Health Study survey instrument race would be reduced, or even disappear, ly distributed (Chantala and Tabor, 1999). did not specifically ask juveniles about vic- if adequate controls for neighborhood 3. The minority racial category (about 45 timization in the home (e.g., child abuse) type were included in the model. percent of the sample) includes the fol- or about property victimization. lowing groups: African Americans (22 per- 8. In this study, telescoping could have oc- cent of the sample); Asians (6 percent), curred only for year 1, because the infor- References American Indians (2 percent), Hispanic mation for year 2 was bounded by the Bard, M., and Sangrey, D. 1986. The Crime whites (7 percent), and other racial or interview for year 1. In the first interview, Victim’s Book, 2nd ed. New York, NY: ethnic groups (8 percent). even though respondents were asked about Bruner/Mazel Publishers. 4. For a large group of respondents, the events that occurred during the prior 12 Chantala, K., and Tabor, J. 1999. Strategies second interview took place less than 12 months, they might have reported events To Perform a Design-Based Analysis Using months after the initial interview. Because that occurred before that period. Examina- the Add Health Data. Chapel Hill, NC: Caroli- the majority of the measures used in the tion of research with the NCVS data has na Population Center, University of North analyses asked respondents about “the suggested that victimization rates were Carolina at Chapel Hill. Copies may be last 12 months,” those whose second inter- higher for unbounded interviewees (those viewed or printed at www.cpc.unc.edu/ view took place less than 11 months after who were not interviewed 6 months earli- projects/addhealth/strategies.html. the initial interview were excluded from er) than for bounded interviewees (those the analyses. who were interviewed 6 months earlier) Cohen, L.E., and Felson, M. 1979. Social (Murphy and Cowan, 1982). change and crime rate trends: A routine 5. Some households had more than one 9. The physical development criteria list- activity approach. American Sociological child represented in the sample. To elimi- Review 44(4):588–609. nate the bias that would otherwise result ed on page 5 were standardized, used from the fact that responses from children to create separate scales for males and Elliott, D.S., Ageton, S.S., Huizinga, D., within the same household were correlat- females, and then recombined into a sin- Knowles, B.A., and Cantor, R.J. 1983. The ed, only the youngest child in each house- gle scale. Juveniles who scored in the 50th Prevalence and Incidence of Delinquent hold was included in the analyses. If a percentile or higher were coded as more Behavior: 1976–1980: National Estimates of household had two children of the same physically developed, and those who Delinquent Behavior by Sex, Race, Social age, one was randomly selected for the scored below the 50th percentile were Class, and Other Selected Variables. Boul- analyses. coded as less physically developed. der, CO: Behavioral Research Institute. 6. All of the analyses and statistical tests 10. The analysis of relationships between Esbensen, F., and Huizinga, D. 1991. Juve- for this Bulletin were conducted using the violent offending and victimization for the nile victimization and delinquency. Youth survey estimation procedures in Stata four categories of drug use also explored & Society 22(2):202–228. Release 7, a publicly available statistical variations by age, gender, and race. Satu- package capable of handling complex sur- rated models, which included all possible vey designs such as that of the Add interactions between these three variables Health Study.

9 Fagan, J., Piper, E.S., and Cheng, Y. 1987. Johnston, L.D., Bachman, J.G., and O’Malley, Sampson, R.J., and Lauritsen, J.L. 1994. Contributions of victimization to delin- P.M. 2001. Monitoring the Future: Question- Violent victimization and offending: quency in inner cities. Journal of Criminal naire Responses From the Nation’s High Individual-, situational-, and community- Law and Criminology 78(3):586–609. School Seniors, 1998. Ann Arbor, MI: Insti- level risk factors. In Understanding and tute for Social Research, University of Preventing Violence, vol. 3, edited by A.J. Federal Bureau of Investigation. 2000. Michigan. Reiss and J.A. Roth. Washington, DC: Crime in the United States 1999. Washing- National Academy Press, pp. 1–114. ton, DC: U.S. Government Printing Office. Kilpatrick, D., Best, C., Veronen, L., Amick, A., Villeponteaux, L., and Ruff, G. 1985. Snyder, H.N., and Sickmund, M. 1999. Juve- Finkelhor, D., and Ormrod, R. 1999. Report- Mental health correlates of criminal vic- nile Offenders and Victims: 1999 National ing Crimes Against Juveniles. Bulletin. Wash- timization: A random community survey. Report. Washington, DC: U.S. Department ington, DC: U.S. Department of Justice, Journal of Consulting and Clinical Psycholo- of Justice, Office of Justice Programs, Of- Office of Justice Programs, Office of Juve- gy 53(6):866–873. fice of Juvenile Justice and Delinquency nile Justice and Delinquency Prevention. Prevention. Lauritsen, J.L., Sampson, R.J., and Laub, Hawkins, J.D., Herrenkohl, T.I., Farrington, J.H. 1991. The link between offending and Sparks, R. 1982. Research on Victims of D.P., Brewer, D., Catalano, R.F., Harachi, victimization among adolescents. Criminol- Crime. Washington, DC: U.S. Government T.W., and Cothern, L. 2000. Predictors of ogy 29(2):265–292. Printing Office. Youth Violence. Bulletin. Washington, DC: U.S. Department of Justice, Office of Jus- Murphy, L.R., and Cowan, C.D. 1982. Effects Preparation of this Bulletin was funded by the tice Programs, Office of Juvenile Justice of bounding on telescoping in the National National Center for Juvenile Justice’s National and Delinquency Prevention. Crime Survey. In The National Crime Survey: Juvenile Justice Data Analysis Project, which is Working Papers, vol. II: Methodological Hindelang, M.J. 1981. Variations in sex- supported by cooperative agreement number Studies, edited by R.G. Lehner and W.G. race-age-specific incidence rates of offend- 99–JN–FX–K002 with the Office of Juvenile Skogan, Washington, DC: U.S. Department ing. American Sociological Review 46(4): Justice and Delinquency Prevention, Office of of Justice, Office of Justice Programs, 461–474. Justice Programs, U.S. Department of Justice. Bureau of Justice Statistics, pp. 83–89. Hindelang, M.J., Gottfredson, M.R., and Riggs, D., Rothbaum, B., and Foa, F. 1995. Garofalo, J. 1978. Victims of Personal Crime: The Office of Juvenile Justice and Delinquency A prospective examination of symptoms An Empirical Foundation for a Theory of Prevention is a component of the Office of of posttraumatic stress disorder in victims Personal Victimization. Cambridge, MA: Justice Programs, which also includes the of nonsexual assault. Journal of Interper- Ballinger. Bureau of Justice Assistance, the Bureau of sonal Violence 10(4):201–214. Justice Statistics, the National Institute of Hough, M., and Mayhew, P. 1983. The Sampson, R.J., and Lauritsen, J.L. 1990. Justice, and the Office for Victims of Crime. British Crime Survey: First Report. London, Deviant lifestyles, proximity to crime, and England: Her Majesty’s Stationery Office. the offender-victim link in personal vio- Jensen, G.F., and Brownfield, D. 1986. Gen- lence. Journal of Research in Crime and der, lifestyles, and victimization: Beyond Delinquency 27(2):110–139. routine activity. Violence and Victims 1(2): 85–99.

10 Acknowledgments Jennifer N. Shaffer is a doctoral can- didate at the Pennsylvania State Uni- versity and a predoctoral fellow at the National Consortium on Violence Research. R. Barry Ruback is Pro- fessor of Crime, Law, and Justice and Sociology at the Pennsylvania State University. This research is based on data from ant to know more about the issues the Add Health project, a program W designed by J. Richard Udry (princi- It’s Fast in this Bulletin or related information? pal investigator) and Peter Bearman Log on to ojjdp.ncjrs.org: and funded by grant P01ÐHD31921 from the National Institute of Child ➤ Health and Human Development Browse titles alphabetically or to the Carolina Population Center, It’s Easy by topic. University of North Carolina at Chapel Hill, with cooperative funding partici- ➤ pation by the National Cancer Insti- Discover the latest OJJDP releases. tute; the National Institute of Alcohol Abuse and Alcoholism; the National ➤ Subscribe to OJJDP’s listserv Institute of General Medical Sciences; It’s Free JUVJUST and the electronic the National Institute of Mental Health; the National Institute of Nursing Re- newsletter JUSTINFO. search; the Office of AIDS Research, National Institutes of Health (NIH); ➤ Link to the NCJRS Abstracts Data- the Office of Behavior and Social Science Research, NIH; the Office base to search for publications of the Director, NIH; the Office of of interest. Research on Women’s Health, NIH; the Office of Population Affairs, U.S. Department of Health and Human Services (HHS); the National Center for Health Statistics, Centers for Dis- ease Control and Prevention, HHS; Share With Your Colleagues the Office of Minority Health, Cen- Unless otherwise noted, OJJDP publications are not copyright protected. We ters for Disease Control and Preven- encourage you to reproduce this document, share it with your colleagues, and tion, HHS; the Office of Minority reprint it in your newsletter or journal. However, if you reprint, please cite OJJDP Health, Office of the Assistant Sec- and the authors of this Bulletin. We are also interested in your feedback, such as retary for Health, HHS; the Office of how you received a copy, how you intend to use the information, and how OJJDP the Assistant Secretary for Planning materials meet your individual or agency needs. Please direct your comments and and Evaluation, HHS; and the Na- questions to: tional Science Foundation. Juvenile Justice Clearinghouse Publication Reprint/Feedback P. O. Box 6000 Rockville, MD 20849Ð6000 800Ð638Ð8736 301Ð519Ð5600 (fax) E-mail: [email protected]

11 U.S. Department of Justice PRESORTED STANDARD Office of Justice Programs POSTAGE & FEES PAID Office of Juvenile Justice and Delinquency Prevention DOJ/OJJDP *NCJ~195737* PERMIT NO. GÐ91 Washington, DC 20531 Official Business Penalty for Private Use $300

Bulletin NCJ 195737