Trends in Juvenile Violent Offending: an Analysis of Victim Survey Data

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

J. Robert Flores, Administrator OctoberMarch 2002

Trends in Juvenile Violent A Message From OJJDP Offending: An Analysis The fact that our assessment of current trends in juvenile offending of Victim Survey Data is based largely on arrest data reported to the Federal Bureau of Investi- gation (FBI) by local law enforcement agencies raises a fundamental ques- James P. Lynch tion about the capacity of such data to provide an accurate and compre- Current understanding of juvenile offend- understanding of juvenile offending reflects hensive picture of the myriad chal- ing nationally comes primarily from the ar- more the nature of crime or the peculiari- lenges that face today’s youth and rest rates compiled by the Federal Bureau ties of the filters through which the prob- our society. of Investigation (FBI) for its annual Uni- lem is viewed. form Crime Reports (UCR) (Snyder and The information that this Bulletin of- Sickmund, 1999; Snyder, 2000). This Bul- fers on trends in juvenile violent offending over the past two decades, letin uses data from the National Crime Trends in Arrest Rates however, comes from a different Victimization Survey (NCVS) to estimate From the FBI’s Uniform source: the victims of those offenses. offending rates for juveniles from 1980 Crime Reports through 1998 from another perspective. Thus, unlike the data derived from According to UCR data, on average, for the the FBI’s Uniform Crime Reports, The juvenile offending rates derived from period 1980 to 1998, the juvenile arrest which drive traditional assessments, the NCVS data are examined by gender, rate for serious violent crimes was 5 per the information provided by the Na- race, and type of crime. These trends are 1,000 persons ages 12 to 17 in the residen- tional Crime Victimization Survey— compared with arrest rate trends based on tial population of the United States.1 Ar- and featured in these pages—is not UCR data. Because victims have little in- rests decreased during the early 1980s and limited to cases that come to the at- formation on crimes in which they do not then (with the exception of 1987) increased tention of local law enforcement offi- have face-to-face contact with offenders, steadily from 1984 until 1994 (figure 1). cials, primarily through arrests. this analysis is restricted to violent offend- They began to decrease in 1995 and con- By comparing the pictures of trends ing. It is further restricted to relatively se- tinued to decrease through 1998. in juvenile violent offending that these rious violent crimes—aggravated assault, Throughout the period 1980 to 1998, the diverse data sources provide, we can robbery, and forcible rape. Simple assaults begin to answer the critical question are excluded because this crime class in- arrest rates for serious violent crimes were much higher for male juveniles than posed above and to determine cludes a large number of very minor inci- whether our present understanding of for female juveniles (figure 2). On average, dents that are infrequently reported to the juvenile offending accurately reflects the arrest rate for male juveniles was 6.6 police and are arguably very different from the nature of those crimes—and not the incidents that are likely to be the ob- times that for female juveniles—8.6 per merely the nature of its origins. ject of police investigation. Unlike UCR 1,000 and 1.3 per 1,000, respectively.2 The data, NCVS data are not limited to cases overrepresentation of males in the juvenile Accordingly, it is our hope that exam- that come to the attention of police or re- arrest rates for serious violent crimes ining information from a variety of sult in arrests. Comparing the pictures that decreased over time. In 1980, the arrest sources about a variety of activities related to juvenile offending will better emerge from these different data sources rate for male juveniles was 8.4 times the equip us to prevent and combat such can help determine whether the current rate for female juveniles. This imbalance delinquent acts. Access OJJDP publications online at ojjdp.ncjrs.org Figure 1: Juvenile Arrest Rates for Serious Violent Crimes: 1980Ð98

1 Arrests (per 1,000 juveniles) 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Source: National Center for Juvenile Justice.

Figure 2: Juvenile Arrest Rates for Serious Violent Crimes, by Gender: 1980Ð98

12 10 8 6 4 2

0 Arrests (per 1,000 juveniles) 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Male Female

Source: National Center for Juvenile Justice. declined steadily so that by 1998 the arrest black juveniles in arrests for serious vio- Limitations of Police- rate for male juveniles was only 4.5 times lent crimes decreased over time. In 1980, that for female juveniles. the arrest rate for black juveniles was 6.1 and Offender-Based times the rate for white juveniles, and in Data Sources Throughout the period, the arrest rates 1981, the ratio was 6.6. By 1998, it had for serious violent crimes were much One must be cautious in using the forego- fallen to 3.9. Although this decrease was ing arrest data to make statements about higher for black juveniles than for white pronounced, it was not consistent through- juveniles (figure 3)—on average, 16.6 per the offender population and the offending out the period. Much of the decrease oc- rate. UCR arrest data include a highly se- 1,000 and 3.0 per 1,000, respectively. The curred between 1996 and 1998. disproportionately larger involvement of lect group of offenders. Being arrested is

2 Figure 3: Juvenile Arrest Rates for Serious Violent Crimes, by Race: 1980Ð98

25 20 15

10 Arrests 5

(per 1,000 juveniles) 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Ye a r

White Black

Source: National Center for Juvenile Justice. imperfectly related to committing an of- offenders are likely to be overrepresented Trends in Offending fense. Fewer than 50 percent of all violent in UCR arrest data. crimes are reported to the police (Renni- Rates From the The picture of the offender population son, 2000). Even when crimes are reported, National Crime emerging from self-report surveys of crim- many offenders are never apprehended inal behavior is also somewhat suspect. Victimization Survey (Federal Bureau of Investigation, 1999). The samples used in these surveys are The NCVS has existed since 1973 and can, Of those individuals identified by the po- relatively small, and there is some con- therefore, be used to assess offending lice as offenders, many are not formally cern that, because certain segments of rates from that time until 1998 (the year charged. Even if police arrested every per- the population—specifically, minority for which the most recent data are avail- son who committed a crime, UCR data males—are not adequately captured in able). However, because the methodology would still be limited because almost 35 these household surveys, their represen- employed in the survey changed radical- percent of all police departments in the tation in the offending population will be ly in 1992, the pre-1992 data used in this United States fail to completely report understated. Finally, self-report surveys analysis have been adjusted to make them their arrest data to the FBI (Jarvis and of offending are susceptible to response comparable to the post-1992 data (see Lynch, 1996; Maltz, 1999). This nonreport- errors in which respondents strategically “Adjusting NCVS Data” on page 16). ing of arrest data is not uniform across the overstate or underreport their offending country. Larger jurisdictions report more behavior. completely than smaller jurisdictions. This Advantages and filtering of cases by the police and police- For these reasons, the descriptions of the Disadvantages of NCVS Data based data systems can introduce sub- offending population that emerge from The NCVS has some distinct advantages as stantial bias in the data. these data sources are suspect. Therefore, a source of data on offending behavior, at it is useful to examine the NCVS as anoth- For these reasons, using arrest rates to least for serious violent offenses (Planty, er source of information on offenders and approximate offending rates will undoubt- 2000): offending rates. Although the NCVS has edly produce low estimates of offending. ◆ limitations of its own that call the resulting First, the NCVS should capture a greater To the extent that attributes of offenders data into question, these limitations are percentage of offending behavior than (e.g., race, gender) affect the probability different from those affecting data gathered can be captured in police records. A of being arrested, UCR arrest data may from police reports and self-report surveys large, nonrandom portion of offending overrepresent “arrest-prone” groups of of offending behavior. Consequently, the behavior is not reported to the police. offenders. In addition, local police depart- NCVS can be used in concert with these Some unknown proportion of offending ments’ different rates of reporting to the traditional sources to paint a more com- behavior is not reported in victim sur- FBI can also distort the picture of the of- plete picture of the offender population. If veys, but it is certainly less than the fender population. If large urban police characteristics of the offender population proportion not reported to the police. departments that service areas with rela- are similar across the three data sources, ◆ Second, the NCVS uses larger samples tively large black populations, for exam- then it is more likely that these statistical than the National Longitudinal Survey of ple, report more completely than smaller systems are accurately describing offend- Youth (NLSY) and other large-scale self- suburban departments that service pre- ers and offending trends. To the extent that report surveys of offending behavior. dominately white communities, then black the data sources provide varying pictures This larger sample should increase the of juvenile offending, an attempt must be reliability of the estimates of offending. made to understand why.

3 Computing Offending Rates With the NCVS The National Crime Victimization Survey Weighting Incident Rates To Victims could not report the age of the (NCVS) is designed to produce victimi- Produce Offending Counts offenders in only 2 percent of the inci- zation rates and incident rates for the dents of violence reported in 1998. Non- Generating an offending rate requires an residential population of the United response for gender and race of the estimate of the number of offending events States based on a sample from that offender was only 0.2 percent and 0.9 (that is, crimes committed), which is divid- population: percent, respectively. Because so few ed by the population being studied. A data are missing (and because several ◆ given crime incident can have as many Victimization rates estimate the attempts to impute these data proved offending acts as there are offenders in- number of people in a given popula- unsuccessful), cases with missing involved. An incident involving three offend- tion who have been victims. If two formation on age, race, or gender were ers, then, would contribute three offend- people are robbed at gun point, this excluded from this analysis. This will ing acts to the offending rate because would contribute two victimizations to lead to a slight underestimate of the three different people committed the crim- the victimization rate. If four people offending rate. were robbed, then four victimizations inal act. A crime perpetrated by only one person would contribute one offense to would be added to the victimization Co-Offending the offending rate.3 The number of offend- rate. The victimization rate is the sum ing acts is computed by multiplying the The problem of mixed age, race, and of the estimated number of victimiza- incident weight by the number of offend- gender groups of offenders is more se- tions divided by the estimate of the ers present during the criminal event. rious than the missing data problem population. because a much larger proportion of crimes involve co-offending than miss- ◆ Incident rates indicate the number Complications in Computing ing data. From an information point of of crime events that have occurred. Offending Rates With view, it would be best to ask the re- The robbery with two victims would the NCVS spondent about each offender, but this contribute only one event to the in- Calculating the offending rate with the is burdensome for the respondent and cident rates because the two peo- NCVS is complicated by nonresponses could increase nonresponse in the sur- ple were robbed in the same event. resulting in missing data, by the way in vey. The NCVS attempts to balance the To avoid the possibility of double- which the survey asks about crimes in- requirements of maintaining the quality counting an event with two victims, volving co-offending, and by the “series of information and limiting the burden any victimization is divided by the incident” procedure used to accommodate placed on respondents by asking about number of victims in the incident, high-volume repeat victimization. It is im- groups of offenders collectively. So, because each victim has the poten- portant, therefore, to describe how these the survey includes questions about tial to be in the sample and report problems were taken into account in the whether all of the assailants were of the event.1 computation of the offending rate. the same race or gender and what that race or gender was. The problem with Rates can be computed for the entire reporting on aggregates of offenders is 2 Missing Data population or for specific subgroups. that there will be mixed groups that can- Nonresponse or missing data are a prob- not be allocated to a given race, gender, lem in any data set. Missing data are infre- or age group. If the group includes two quent in the NCVS, compared with most 1 The robbery with two victims would be divided by 2 men and two women, all that can be other data collections (e.g., the UCR or (i.e., one event/two victims) and the result (0.5) would known from the NCVS data is that the be multiplied by the sample weight to obtain the con- National Woman’s Study). The NCVS has group contained both men and women. tribution of this event to an incident count. The results a 95-percent response rate and very low for all incidents would be summed across all respond- item nonresponse. In the case of informa- Age. Because juvenile offending is the ents to get an incident count for the population. This tion on offenders in violent crime incidents, focus of this study, it is important to ac- incident count would be divided by the weighted pop- the amount of missing data is small. curately characterize offenders by age. ulation count, as in the case of victimization rates, to obtain an incident rate for the population. Moreover, the amount of adult and juvenile co-offending is substantial. For these 2 A subgroup’s estimated number of victimizations is 3 For a very small proportion of serious violent incidents, reasons, mixed-age groups could not developed by multiplying the subgroup’s reported num- the number of offenders reported by the respondent is simply be omitted from the analysis. In- ber of victimizations by the subgroup’s sample weight. implausibly large. For the years 1993 to 1998, for exam- stead, this Bulletin provides two esti- The subgroup’s population is determined by multiplying ple, victims reported between 1 and 96 offenders pres- mates of juvenile offending. Offending the number of respondents in the subgroup by the sub- ent during an incident. Moreover, approximately 1 per- rates that included adult co-offending group’s sample weight. The subgroup’s sample weight cent of the serious violent incidents involved more than 10 offenders. Given the fact that these rare events can assumed that all offenders in mixed-age is the inverse of the probability of being selected into groups of offenders were juveniles. Of- the sample. If 10 people are sampled from a subgroup contribute a great deal to the offending rate and that a of 100, the probability of selecting a given person from respondent cannot accurately count offenders beyond fending rates that excluded adult co- that subgroup is 1 over 10 or 10 percent. The inverse a certain number, the computations of offending rates offending assumed that all offenders in of the probability (i.e., 1/0.1) is equal to 10. To produce presented in this Bulletin counted all events with 10 or mixed-age groups of offenders were subgroup population estimates from their sample more offenders as having 10 offenders. adults. The first of these estimates will counts, the counts from the sample would be multiplied by the inverse of the probability of selection, or 10.

4 be too high and the other will be too low, the data too complex and the omission of in series incidents is too high, but there with a reasonable estimate of the juve- the mixed groups data did not make as is also reason to believe that some of nile offending rate being somewhere be- much difference in these rates as it did in these reports of high-volume repeat tween the two. In this sense, the two es- the age-specific rates. victimization are accurate. Including se- timates (including and excluding adult ries incidents as one crime probably co-offending) provide an interval within Series Incidents contributes to an underestimate of the which the actual juvenile offending rate A series incident refers to a procedure juvenile offending rate, but the amount 4 most likely falls. developed by the Census Bureau and it contributes is unclear. Gender and race. A different approach used in the NCVS that reduces the burden In sum, the procedures used to take ac- was taken to estimating gender- and of collecting information on high-volume count of missing data, co-offending, and race-specific rates of juvenile offending. repeat victimization. When a respondent series incidents all contribute to the un- Here, mixed-gender and mixed-race reports he or she was victimized more derestimation of offending rates. The age- groups that had a majority of one race or than five times in the past 6 months, when specific rates are least affected by the one gender were treated as if all mem- these events are similar, and when the procedural peculiarities of the NCVS be- bers of the group were of that race or respondent cannot report on the details cause all mixed-group offending was in- gender. So, if the group included one of each, then the interviewer is instructed cluded in these estimates. As a result, male and three females, all offenders to note the number of incidents but only for the age-specific rates, only missing were counted as female. When the re- collect detailed information on the most data and the handling of series incidents spondent reported that no race or gen- recent event. Because the Bureau of contribute to the underestimate. The der group was in the majority, the case Justice Statistics, which sponsored the race- and gender-specific rates are more was omitted from the estimates. Approx- NCVS, is concerned about the accuracy substantial underestimates because imately 7 percent of the violent incidents of the information provided and, specifi- some mixed-offending groups (those fell into this category. Here again, this cally, the number of events reported to perfectly balanced in terms of race and estimation method will contribute to un- have occurred, it excludes these events gender) are omitted. This should not af- derestimating the juvenile offending rate. from annual rate estimates. fect comparisons of rates across race There is no reason to believe, however, The estimates in this Bulletin include se- and gender groups. that mixed-gender or mixed-race groups ries incidents, but each one only as one occur more often in a particular race or incident and not as the number of inci- gender group and that omitting them dents that the respondent said occurred 4 The vast majority of series incident victimizations would lower the rates of one group rela- in the series. Although series incidents involve domestic violence and crimes at work (e.g., tive to another. Hence, comparisons assaults and thefts that occur at work) (Dodge, 1987; are a small part of annual victimization Lynch, Berbaum, and Planty, 1999). These events sel- of the race- and gender-specific rates rates, when they are weighted by the should be appropriate. The strategy of dom involve juveniles. Approximately 12 percent of number of individual incidents occurring series incidents take place at school, and these are making multiple estimates was not fol- in them, they become more consequen- more likely to involve juveniles (Lynch, Berbaum, and lowed for gender- and race-specific rates tial. There is reason to believe that the Planty, 1999). because it would make presentation of number of events reported as occurring

◆ Third, some reason exists to believe self-interest for strategic response in race, sex, and number of offenders and that the response errors in the NCVS self-report surveys of offending. about their relationship to the offender. Surveys of offenders, obviously, can may be less severe and better under- ◆ Finally, the NCVS includes a great deal of stood than those in self-report surveys provide much more detailed informa- detailed information on individual crime tion about the offender. of offending. Respondents to self-report events. In contrast, surveys of offend- surveys of offending may have substan- ing behavior are less concerned with ◆ Second, some researchers question tial reasons for not responding accurate- describing individual crime incidents the ability of victims to provide reliable ly or truthfully—they may be concerned than with estimating the prevalence of information on offenders, even in vio- about repercussions or may want to con- offending over a given period of time. lent crimes. Although the race and gen- vey a certain image (Thornberry and This information on the social context der of offenders may be relatively easy Krohn, 2000). The victim does not have of offending (e.g., where the crime to identify, the same is not the case for the same self-interest in reporting about occurred, who was present, the num- age—for example, it may be difficult to the attributes of offenders. To be sure, ber of offenders) can provide a descrip- make the distinction between a 21-year- there are a number of known response tion of offending that is simply not old adult and a 17-year-old juvenile. errors in the NCVS that influence the available in existing surveys of offend- ◆ Third, design features of the NCVS may reporting of victimization events, and ing behavior.3 also limit its utility for describing the incomplete reporting of crimes will offender population accurately. The lead to distortions in the description The NCVS also has some clear limitations fact that the NCVS only interviews of offenders (Biderman and Lynch, for estimating offending rates: victims 12 years of age and older will 1991); however, it seems likely that esti- ◆ First, the NCVS asks very few questions probably result in an underestimate of mates of offending are affected less by about the attributes of offenders. Re- the offending rate for juveniles, espe- these sources of error than by the clear spondents are asked about the age, cially the offending rate for children

5 younger than 12 who are most likely violent crimes was, on average, somewhere dropped to around 31 per 1,000 in 1982, to prey upon younger victims (Singer, between 44 and 75 victimizations by juve- where it hovered through 1987. In 1988, it 1981). However, because these younger nile offenders per 1,000 persons ages 12 to increased to 36 per 1,000, then increased victims are often victimized by adult 17 in the U.S. population. The lower esti- sharply a few years later, reaching 91 per caretakers (e.g., in sex offenses), inter- mate excludes all co-offending with adults, 1,000 in 1993. Thereafter, the rate declined viewing these child victims could also while the higher estimate includes all of- to 36 per 1,000 in 1998. increase the proportion of older offend- fenses in which there was co-offending When adult co-offending incidents are ex- ers recorded in the data. with adults (see discussion of co-offending amined separately, a very sharp increase in in “Computing Offending Rates With the ◆ Fourth, the NCVS asks about only a adult co-offending between 1987 and 1988 NCVS” on pages 4–5). Over the period narrow range of crimes, and it can only is revealed (71 percent). This compares 1980 to 1998, almost half the offending by produce offending rates for those types with a 16-percent increase in juvenile-only juveniles was done in concert with per- of crime. offending. The adult co-offending rate sons perceived by victims to be adults.4 ◆ Fifth, the NCVS cannot be used to pro- dropped dramatically from 1991 to 1992 duce prevalence estimates of offending From a high in 1981, the offending rate (a 38-percent decrease), while the juvenile- behavior because offenders cannot be (whether including or excluding adult co- only rate increased. This suggests that linked across crime incidents. offending) decreased until approximately adult co-offending contributed dispropor- 1988, when it began to increase sharply tionately to the increases in total juvenile Readers should keep these strengths and and stayed relatively high until 1994. At offending in the late 1980s but did not weaknesses in mind as they consider the that point, it decreased sharply through drive the decreases in total juvenile of- offending rates presented in the following 1998 (figure 4). fending in the mid-1990s. While adult co- sections. offending dropped dramatically between Including adult co-offending, the offending 1991 and 1992, this decrease was offset by rate for serious violent crimes was 80 per Trends in Offending Rates increases in juvenile-only offending. It was 1,000 in 1981. It dropped to 56 per 1,000 Since 1980 not until later in the decade (approximate- by 1987 but then rose sharply to 78 per ly 1994) that juvenile-only offending began The offending rates presented here are es- 1,000 in 1988.5 It remained near this level to decrease markedly. timated with data from the NCVS. Overall until 1992, when it rose to 89 per 1,000, trends are reported along with separate later peaking at 120 per 1,000 in 1993. In Offending rates since 1980, by gender. trends by gender and race. Gender and 1994, the rate began to decline until it Over the period 1980 to 1998, the average race were chosen for separate considera- reached 51 per 1,000 in 1998. offending rate for serious violent crimes tion because these are the only attributes was 8 to 11 times higher (depending on The contours of the trend are reasonably of offenders included in both the UCR and whether adult co-offending is excluded or similar when co-offending with adults is the NCVS. included) for male juveniles than female excluded. From 44 per 1,000 in 1981, the juveniles. Excluding adult co-offending, the Offending rates since 1980, by adult juvenile offending rate (excluding adult average offending rates for male juveniles co-offending status. In the period 1980 to co-offending) for serious violent crimes 1998, the juvenile offending rate for serious and female juveniles were 68 per 1,000 and

Figure 4: Juvenile Offending Rates for Serious Violent Crimes, by Adult Co-Offending Status: 1980Ð98

120

100

Offenses (per 1,000 juveniles) Offenses 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Ye a r

Without adult co-offending With adult co-offending

Sources: U.S. Department of Justice, Bureau of Justice Statistics, National Crime Surveys: National Sample 1979Ð87 (computer file); National Crime Surveys: National Sample 1986Ð91 (computer file); National Crime Victimization Survey 1992Ð98 (computer file).

6 8 per 1,000, respectively. Including adult The trends in the female offending rate more likely than female juveniles to engage co-offending, the offending rate for male (either including or excluding adult co- in serious violent co-offending with adults. juveniles was, on average, 122 per 1,000 offending) exhibit the same general move- compared with 11 per 1,000 for females. ments from highs in the first half of the The overrepresentation of males in serious Figure 5 shows the juvenile offending rates 1980s to lows in the second half of the violent offending decreased over time. for serious violent crimes by adult co- 1980s to highs in the early 1990s and de- Excluding adult co-offending, from 1980 to offending status and gender for the years clines thereafter. These movements are 1986, male juveniles, on average, were en- 1980 to 1998. more erratic than the trends in male of- gaged in serious violent offending at 16 fending, in large part because the female times the rate of female juveniles. In the Including adult co-offending, the offending rate is so low in general. period 1993 to 1998, this ratio decreased rate for male juveniles was high in 1981 to 8. This decrease resulted more from (140 per 1,000) and declined unevenly to The ratio of male-to-female offending rates an increase in the offending rate of female a low of 88 per 1,000 in 1985. The rate in- for serious violent crimes is greater when juveniles than a decrease in the offending creased slowly in 1986, dramatically in- adult co-offending is included than when it rate of male juveniles. creased to 135 per 1,000 in 1988, and is excluded. On average, during the period, reached a high of 181 per 1,000 in 1993. the total offending rate (including adult co- Offending rates since 1980, by race. Thereafter, it declined steadily to a low of offending) for male juveniles was 11 times Over the period 1980 to 1998, the average 83 per 1,000 in 1998. These trends are simi- that for female juveniles. When adult co- offending rate for serious violent crimes lar when co-offending with adults is exclud- offending is excluded, the ratio decreases was 3.4 to 4.1 times higher (depending on ed from the rates. to 8. This suggests that male juveniles are whether adult co-offending is excluded or included) for black juveniles than white

Figure 5: Juvenile Offending Rates for Serious Violent Crimes, by Adult Co-Offending Status and Gender: 1980Ð98

Male Juvenile Offending 200

160 120

Offenses Offenses 40

(per 1,000 juveniles) 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Female Juvenile Offending 25

Offenses Offenses 5

(per 1,000 juveniles) 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Without adult co-offending With adult co-offending

7 juveniles. The average offending rate for Adult co-offending accounted for a dispro- Trends in Offending Rates black juveniles was between 96 per 1,000 portionate amount of the increase in seri- Since 1993 (excluding adult co-offending) and 167 ous violent offending for black juveniles The longer term trends reported in the per 1,000 (including adult co-offending) in the late 1980s (figure 6). Rates of adult previous section were limited to broad compared with between 28 and 41 per co-offending for blacks increased by 60 categories of gender and race because 1,000, respectively, for white juveniles. percent between 1986 and 1987 and by 37 the adjustments that must be applied to percent from 1988 to 1989. The juvenile- Adult co-offending was a more prominent pre-1992 NCVS data are more reliable for only offending rates for blacks during this feature of serious violent offending for larger groups than smaller ones. In the period increased much less (increasing 15 black juveniles than for white juveniles. period 1993 to 1998, no adjustments need percent from 1986 to 1987 and decreasing On average, over the period 1980 to 1998, to be applied, so it is possible to examine 2 percent between 1988 and 1989). Later, the ratio of adult co-offending to total juve- more specific demographic and offense between 1991 and 1993, this trend was re- nile offending was 0.31 for whites (approx- categories. Moreover, because some of versed as the increases in juvenile-only imately 1 instance of adult co-offending the more pronounced decreases in ar- offending were larger than the increases in for every 3 instances of juvenile offending) rests and offending occurred in this pe- adult co-offending.6 and 0.43 for blacks (more than 2 instances riod, a more detailed examination may of adult co-offending for every 5 instances be enlightening. of juvenile offending).

Figure 6: Juvenile Offending Rates for Serious Violent Crimes, by Adult Co-Offending Status and Race: 1980Ð98

Black Juvenile Offending 300 250

200 150

Offenses Offenses 100 50 (per 1,000 juveniles) 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

White Juvenile Offending 60 50 40 30

20 Offenses Offenses 10 (per 1,000 juveniles) 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Without adult co-offending With adult co-offending

8 Offending rates since 1993, by specific 1,000 in 1993 to 30.9 per 1,000 in 1998, or crime category. The decrease in serious 54 percent. When adult co-offending is ex- An Important Distinction violent offending between 1993 and 1998 cluded, the rate decreased from 40.9 to When interpreting the rates presented was greatest for robbery and, in some 14.7, or 64 percent. Therefore, for both in this Bulletin, it is important to ap- cases, less for aggravated assault (figure robbery and aggravated assault, with 7 preciate the difference between of- 7). In 1993, the total juvenile offending adult co-offending included or excluded, fending rates and offender rates. An rate (including adult co-offending) for rob- the rate of serious juvenile violence in offending rate estimates the number bery was 46.6 per 1,000. This fell to 14.8 1998 was just one-third of its 1993 level. of offenses committed by a given pop- per 1,000 by 1998, a decrease of 68 per- These declines in offending rates for rob- ulation. The offending rate is often recent. When adult co-offending is excluded, bery and aggravated assault are greater ferred to as the incidence rate. An of- the rate decreased from 30.4 per 1,000 in than the corresponding declines in the fender rate estimates the number of 1993 to 10.7 per 1,000 in 1998, a decrease juvenile arrest rates for robbery (41 per- persons in a given population who of 64 percent. The total juvenile offending cent) and aggravated assault (16 percent). have committed one or more crimes. rate (including adult co-offending) for ag- The offender rate is often referred to gravated assault decreased from 67.4 per as the prevalence rate. The difference between these rates is that prevalence rates do not take account of repeated offending by the Figure 7: Juvenile Offending Rates for Robbery and Aggravated same person, whereas incidence rates Assault, by Adult Co-Offending Status: 1993Ð98 do. Once a person has committed one offense, that person is counted in the numerator of the offender rate and Robbery subsequent offending by that person will not affect the rate. In contrast, the 50 numerator of the offending rate will 40 increase whenever a new offense is committed by a member of the at-risk 30 population. One of the advantages of a prevalence rate is that it cannot ex- 20 ceed 1.0 and provides an easily inter- pretable indicator of how many of the Offenses Offenses 10 persons in a given group have offended in a given period of time. With a (per 1,000 juveniles) 0 prevalence rate, one can say, for ex- 1993 1994 1995 1996 1997 1998 ample, that 40 percent of male juveniles have offended in a year. An inci- Year dence rate, on the other hand, can exceed 1.0 if some members of the at-risk population commit more than Aggravated Assault one offense during the period. Inci- 80 dence rates indicate the relative involvement of groups in offending or 70 the relative contribution of a group to 60 total offending while taking into account the size of the group. With an 50 incidence rate, it is possible to say, for example, that males are twice as 40 likely to engage in offending behavior as females, but impossible to say Offenses Offenses 30 what proportion of each group en- 20 gages in offending behavior. (per 1,000 juveniles) 10 Most of the indicators used in criminal justice are incidence rates, largely be- 0 cause it is difficult to identify repeat 1993 1994 1995 1996 1997 1998 offenders or repeat victims, and doing so is essential for estimating preva- Year lence rates. So the UCR crime rates published annually by the FBI are in- Without adult With adult cidence rates, as are the victimization co-offending co-offending rates published by the Bureau of Jus- tice Statistics. It is important to keep Sources: U.S. Department of Justice, Bureau of Justice Statistics, National Crime Victimization in mind the difference between inci- Survey 1992Ð98 (computer file). dence and prevalence rates.

9 Offending rates since 1993, by race and juveniles decreased 76 percent (77 per- contribution of co-offending to the total crime category. The decrease in overall cent excluding adult co-offending). The juvenile offending rate for serious violent serious violent offending was greater for decrease in the offending rate (including crimes. Throughout the period 1993 to black juveniles than for white juveniles adult co-offending) for white juveniles 1998, the average annual juvenile offend- from 1993 to 1998, and this pattern was who committed aggravated assault was ing rate was 87 offenses per 1,000. Lone replicated for both robbery and aggravated 49 percent (50 percent excluding adult offending accounted for 18 offenses per assault (figure 8). For robbery, including co-offending) compared with 70 percent 1,000, co-offending with other juveniles adult co-offending, the offending rate for (79 percent excluding adult co-offending) accounted for 42 per 1,000, and co- white juveniles decreased 46 percent for black juveniles. offending with adults accounted for 27 (45 percent when adult co-offending is ex- per 1,000. Overall, three-fourths of juvenile Trends in co-offending since 1993. It is cluded) and the offending rate for black offending was co-offending, and more than possible, using NCVS data, to examine the

Figure 8: Juvenile Offending Rates for Robbery and Aggravated Assault, by Adult Co-Offending Status and Race: 1993Ð98

White Juvenile Offending 40

10 5 Offenses (per 1,000 juveniles) Offenses 0 1993 1994 1995 1996 1997 1998 Year

Black Juvenile Offending 140 120 100 80

60 40 20

Offenses (per 1,000 juveniles) Offenses 0 1993 1994 1995 1996 1997 1998 Year

Robbery (including adult co-offending) Aggravated assault (including adult co-offending) Robbery (excluding adult co-offending) Aggravated assault (excluding adult co-offending)

Sources: U.S. Department of Justice, Bureau of Justice Statistics, National Crime Victimization Survey 1992Ð98 (computer file).

10 Decreases in serious violent offending from Figure 9: Juvenile Offending Rates for Serious Violent Crimes, by 1993 to 1998 were greater for female juve- Co-Offending Status: 1993Ð98 niles (68 percent) than for male juveniles (55 percent). The decreases in lone offending and adult co-offending were slightly 80 greater for female juveniles (46 and 61 per- 70 cent respectively) than male juveniles (42 and 59 percent, respectively).8 The largest 60 differences between the genders occurred in co-offending with other juveniles, where 50 the decrease for male juveniles was 62 percent compared with a 92-percent decrease 40 for female juveniles. Figure 10 shows the 1993–98 rates of lone offending, juvenile 30 co-offending, and adult co-offending for male and female juveniles. 20 Between 1993 and 1998, lone offending rep- 10

Offenses (per 1,000 juveniles) Offenses resented a much higher proportion of the offending rate for white juveniles than for 0 1993 1994 1995 1996 1997 1998 black juveniles (table 2). Thirty-one percent of the white juvenile offending rate for seri- Year ous violent crimes involved lone offending, 37 percent involved co-offending with adults, and 32 percent involved co-offending with Lone offending Co-offending with juveniles Co-offending with adults juveniles. In contrast, only 20 percent of Sources: U.S. Department of Justice, Bureau of Justice Statistics, National Crime Victimization the black juvenile offending rate for seri- Survey 1992Ð98 (computer file). ous violent crimes involved lone offending, 39 percent involved co-offending with adults, and 41 percent involved co- offending with juveniles. Table 1: Average Juvenile Offending Rates for Serious Violent Crimes, Between 1993 and 1998, the rate of serious by Gender and Co-Offending Status: 1993Ð98 violent offending decreased 72 percent for black juveniles and 40 percent for white Offending Rate (per 1,000 juveniles) juveniles. Lone offending dropped by about the same proportion for both races (42 Committed Committed Committed All percent for black juveniles and 43 percent Gender Alone With Adult With Juvenile Incidents for white juveniles). Co-offending with Male 30.3 55.4 47.7 133.3 adults decreased more for black juveniles Female 4.7 4.3 6.8 15.8 than for white juveniles (63 percent and 50 percent, respectively). The largest difference was for co-offending with juveniles, which decreased 90 percent for black ju- one-third of the co-offending was with co-offending rate decreased more than the veniles and 29 percent for white juveniles. adults. adult co-offending rate (67 percent and 47 Figure 11 shows the 1993–98 rates of lone percent, respectively). All categories—lone offending, co-offending offending, co-offending with juveniles, and with juveniles, and co-offending with Co-offending is somewhat greater among co-offending with adults for white and adults—of the juvenile offending rate for male juveniles than female juveniles (table black juveniles. serious violent crimes decreased during 1). On average, throughout the period 1993 In sum, much of the serious violent offend- 1993–98 (figure 9). The total juvenile of- to 1998, approximately 78 percent of the ing by juveniles is done in groups, and ju- fending rate for serious violent crimes de- male juvenile offending rate for serious vio- veniles are as likely to co-offend with an creased 57 percent, from 119 per 1,000 to lent crimes involved co-offending com- adult as with other juveniles (based on 51 per 1,000. The decrease in offending pared with 70 percent for female juveniles. victims’ perceptions of offender age). Males rates between 1993 and 1998 was greater Co-offending for male juveniles was evenly engage in co-offending more than females, for co-offending than for lone offending. split between adult co-offending (42 percent and blacks more than whites. Between The offending rate for lone offenders who of all incidents) and juvenile co-offending 1993 and 1998, rates of co-offending have committed serious violent crimes de- (35 percent of all incidents). However, for decreased faster than rates of lone offend- creased by 42 percent over the period, female juveniles, a greater proportion of ing. These decreases in co-offending have from 24 per 1,000 to 14 per 1,000. The co- all serious violent crimes involved juvenile been more pronounced for blacks than for offending rate (including both juvenile and co-offenders (43 percent of all incidents) whites. adult co-offenders) declined 61 percent, than adult co-offenders (27 percent of all from 96 to 37 per 1,000. The juvenile incidents).

11 Comparing UCR and Figure 10: Juvenile Offending Rates for Serious Violent Crimes, NCVS Estimates of by Gender and Co-Offending Status: 1993Ð98 Offending Rates As noted previously, UCR data indicate that Male Juvenile Offending the average annual juvenile arrest rate for 90 serious violent crimes during 1980–98 was 5 per 1,000 persons ages 12 to 17. NCVS 80 data indicate that, depending on whether adult co-offending is excluded or included, 70 the juvenile offending rate for the same 60 crimes during the same period was 9 to 15 times the juvenile arrest rate (44 per 1,000 50 excluding adult co-offending and 75 per 1,000 including adult co-offending). This is 40 not unreasonable, considering that only 20 percent of the violent crimes by juve- 30 niles reported in victim surveys are reported to the police and that, of those 20 crimes reported to the police, approximately 50 percent lead to an arrest.

Offenses (per 1,000 juveniles) Offenses 10 The broad trends in the UCR arrest data 0 1993 1994 1995 1996 1997 1998 and NCVS offending data are similar. Both show decreases in the early 1980s followed Year by an increase later in the decade, with continued high rates until the mid-1990s, Female Juvenile Offending when rates dropped dramatically (figure 1 and figure 12). Arrest rates reached a low 18 point in 1987 but increased steadily thereafter until 1994, when they began to drop. 16 Offending rates reached a low point in 14 1987 and then increased in 1988. These higher rates persisted until 1994, when 12 they began to trend downward.

10 The increases observed in the mid-1980s were similar for offending rates and arrest 8 rates, but the decreases in the 1990s were greater for offending rates than for arrest 6 rates. Between 1987 and 1993, the juvenile arrest rate increased about 70 percent, 4 while the juvenile offending rate increased 2 between 114 percent (including adult Offenses (per 1,000 juveniles) Offenses co-offending) and 195 percent (excluding 0 adult co-offending). Regardless of the as- 1993 1994 1995 1996 1997 1998 sumptions that are made about adult co- offending, the increase in the offending Year rate was greater than the increase in the arrest rate. Similarly, the juvenile arrest Lone Offending Co-offending with juveniles Co-offending with adults rate for serious violent crimes declined 27 percent from 1993 to 1998, while the Sources: U.S. Department of Justice, Bureau of Justice Statistics, National Crime Victimization juvenile offending rate for serious violent Survey 1992Ð98 (computer file). crimes declined between 57 percent (including adult co-offending) and 60 percent (excluding adult co-offending) during the same period. The increases and decreases in the offending rate during 1980–98 were much greater than the increases and decreases in the arrest rate.

12 Comparison by Gender Table 2: Average Juvenile Offending Rates for Serious Violent Crimes, The data on both arrests and offending by Race and Co-Offending Status: 1993Ð98 showed higher rates of serious violent offending for male juveniles than female Offending Rate (per 1,000 juveniles) juveniles. The overrepresentation of males Committed Committed Committed All was greater in the data on offending than Race Alone With Adult With Juvenile Incidents in the arrest data. This means that, given their offending rates, male juveniles were White 12.0 14.4 12.2 38.6 underrepresented and female juveniles Black 34.9 68.5 71.6 174.9 were overrepresented in arrest data. The disproportionality of male offending decreased over time but more so in arrest data than in offending data. Figure 11: Juvenile Offending Rates for Serious Violent Crimes, The ratio of male juveniles to female juve- by Race and Co-Offending Status: 1993Ð98 niles was lower in the arrest data than in the offending data. For serious violent White Juvenile Offending crimes (i.e., forcible rape, robbery, and aggravated assault), the average ratio of 20 male juveniles to female juveniles was 18 6.8 in the arrest data but ranged between 9 (excluding adult co-offending) and 16 11.5 (including adult co-offending) in the offending data. This discrepancy may be 14 due to the fact that more crimes involv- 12 ing intimates (e.g., siblings, dating cou- ples) are reported in the NCVS than in 10 police data systems. There may be more 8 male offenders in these crimes than in other types of violent crimes, resulting in 6 a higher ratio of male to female offenders. 4 Domestic violence, for example, is more often committed by males than females. 2 Offenses (per 1,000 juveniles) Offenses Between 1980 and 1998, the male-female 0 ratio for juvenile arrest rates for serious 1993 1994 1995 1996 1997 1998 violent crimes decreased from 8.3 to 4.5 (figure 13). The male-female ratio for ju- Year venile offending rates for serious violent crimes has also generally decreased, but Black Juvenile Offending not as consistently and not as much as the arrest rate ratio. The male-female ratio for 140 offending rates (with adult co-offending 120 included) decreased from a high of 22 in 1981 to 10.5 in 1998 (with a low of 7.3 in 100 1993). When adult co-offending is excluded, the ratio was 16.8 in 1981 and 10.4 in 80 1998 (with a low of 4.9 in 1993). Although this seems broadly consistent with the 60 arrest data, the male-female ratio in offend- 40 ing has fluctuated throughout the period. Indeed, this ratio increased from 1996 to 20 1998 in the offending data, while it continued to decline in the arrest data. Offenses (per 1,000 juveniles) Offenses 0 1993 1994 1995 1996 1997 1998 One of the factors contributing to the decline in the overrepresentation of males Year in juvenile arrest rates for serious violent crimes has been the steady increase in the Lone Offending Co-offending with juveniles Co-offending with adults arrest rates for female juveniles since 1980. Between 1980 and 1998, the female juvenile Sources: U.S. Department of Justice, Bureau of Justice Statistics, National Crime Victimization arrest rate for serious violent crimes in- Survey 1992Ð98 (computer file). creased 89 percent, from 0.9 per 1,000 to 1.7 per 1,000 (figure 2). The arrest rate for

13 Figure 12: Juvenile Arrest and Offending Rates for Serious Violent Crimes, by Adult Co-Offending Status: 1980Ð98

140 120

100 80

60 40 Arrests/Offenses Arrests/Offenses

(per 1,000 juveniles) 20 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Juvenile offending Juvenile offending Juvenile arrests* (including adult co-offending) (excluding adult co-offending)

* Due to the difference in scale between the arrest and offending rates presented in this figure, the rate for juvenile arrests appears relatively flat. Figure 1 presents this juvenile arrest data using a different scale which better shows the fluctuation in juvenile arrest rates during this period. Sources: National Center for Juvenile Justice. U.S. Department of Justice, Bureau of Justice Statistics, National Crime Surveys: National Sample 1979Ð87 (computer file); National Crime Surveys: National Sample 1986Ð91 (computer file); National Crime Victimization Survey 1992Ð98 (computer file).

Figure 13: Male-Female Ratios for Juvenile Arrest and Offending Rates for Serious Violent Crimes, by Adult Co-Offending Status: 1980Ð98

24 20 16 12 8 4 Male-Female Ratio Male-Female 0 1980 1981 1982 1983 1984 1985 1986 1987 19881989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Offending rates Offending rates Arrest rates (including adult co-offending) (excluding adult co-offending)

Sources: National Center for Juvenile Justice. U.S. Department of Justice, Bureau of Justice Statistics, National Crime Surveys: National Sample 1979Ð87 (computer file); National Crime Surveys: National Sample 1986Ð91 (computer file); National Crime Victimization Survey 1992Ð98 (computer file). male juveniles increased only 2 percent the 19 years. In contrast, female offending Comparison by Race during the same period. Moreover, arrest rates varied much more over time and did The data on both offending and arrests rates for male juveniles fluctuated during not show the steady increase found in the showed that compared with white juve- this period, but arrest rates for female ju- female arrest rates (figure 5). niles, black juveniles had much higher veniles increased consistently throughout

14 rates of involvement in serious violent from a high of 5.2 in 1993 to a low of 1.9 Conclusion crime. This disproportionality was greater for 1998, or 63 percent. Examining serious violent offending in arrest rates than in offending rates. through data gathered from victims However, it decreased over time in both In the period 1992–98, the black-white provides a picture that is both similar to the arrest data and the offending data. arrest ratio was quite similar to the black- and different from that emerging from po- Much of the overall decrease in the dis- white offending ratio. On average during lice arrest data. Not surprisingly, the vol- proportionality of serious violent offend- this period, the black arrest rate was 4.9 ume of offending is much greater in the ing by black juveniles was the result of times that of whites. The black offending NCVS data than is indicated in the UCR the large decrease in offending by blacks rate was 4.2 times that of whites when arrest data. Depending on whether adult for these crimes that occurred after 1993. adult co-offending is excluded and 4.6 times that of whites when adult co- co-offending is excluded or included, of- During 1980–98, the average arrest rate for offending is included. fending rate estimates from the NCVS data serious violent crimes was 5.7 times higher are 9 to 15 times greater than the arrest for black juveniles than for white juveniles. The decrease in the overrepresentation of rate estimate from the UCR data. Both The black-white ratio for offending rates blacks in juvenile arrest rates for serious sources of data show race and gender dis- for serious violent crimes ranged from violent crimes resulted in part from the proportionality in serious violent offend- 4.1 (including adult co-offending) to 3.4 steady increase in arrest rates for white ing. Male juveniles offend at much higher (excluding adult co-offending). Although juveniles since the early 1980s. The white rates than female juveniles. Black juve- blacks are overrepresented in both the juvenile arrest rate for serious violent niles offend at much greater rates than offending data and the arrest data for crimes increased from 2.3 in 1980 to 4.2 in white juveniles. The amount of gender 1980–98, the overrepresentation is great- 1994 (figure 3), an increase of 83 percent. disproportionality, however, is greater in er in the arrest data. This rate only began to decrease notice- the NCVS data than the UCR data, while ably in 1996. In contrast, the black juve- the racial disproportionality is greater in Over time, the black-white arrest rate nile arrest rate for serious violent crimes the UCR data than the NCVS data. In other ratio declined from a high of 6.9 in 1983 to increased 64 percent during 1980–94, and, words, victims indicate that there are 3.9 in 1998, or 43 percent (figure 14). The beginning in 1994, it decreased more more male offenders (relative to female comparable offending ratio (including sharply than the arrest rate for white ju- offenders) and more white offenders (rela- adult co-offending) decreased from a high veniles. Looking at offending rates, the tive to black offenders) than appear in of 6.1 in 1992 to 2.4 in 1998, or 61 percent. decrease in the racial disproportionality police records. When adult co-offending is excluded from has resulted from decreases in black of- the offending rates, the ratio decreased fending since 1990, not increases in white offending.

Figure 14: Black-White Ratios for Juvenile Arrest and Offending Rates for Serious Violent Crimes, by Adult Co-Offending Status: 1980Ð98

7 6

4 3

Black-White Ratio Black-White 2 1 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Year

Offending rates Offending rates Arrest rates (including adult co-offending) (excluding adult co-offending)

15 Adjusting NCVS Data As noted in the text, the methodology employed in the Na- By gender. The adjustment ratios for male juvenile offending tional Crime Victimization Survey (NCVS) was changed radi- including adult co-offending are as follows: rape, 3.05; rob- cally in 1992. As a result, data from the pre-1992 period, bery, ns; aggravated assault, ns; simple assault, 1.42; total when the survey was known as the National Crime Surveys violence, 1.41. The adjustment ratios for male juvenile offend- (NCS), must be adjusted to make them comparable to the ing excluding adult co-offending are as follows: rape, ns; rob- more recent data. An adjustment of the data is possible be- bery, ns; aggravated assault, ns; simple assault, 1.69; total cause the old NCS and new (NCVS) designs were run in violence, 1.68. The adjustment ratios for female juvenile of- parallel for 18 months. Data from this overlap period can be fending including adult co-offending are as follows: rape, ns; used to compute the ratio of estimates from the post-1992 robbery, ns; aggravated assault, ns; simple assault, 1.42; total and pre-1992 designs. This ratio, in turn, can be used to ad- violence, 1.41. The adjustment ratios for female juvenile of- just all the pre-1992 data to make them comparable to the fending excluding adult co-offending are as follows: rape, ns; post-1992 data (Kindermann, Lynch, and Cantor, 1997; robbery, ns; aggravated assault, ns; simple assault, 1.69; total Rand, Lynch, and Cantor, 1997; Lynch and Cantor, 1996). violence, 1.68. The Bureau of Justice Statistics (BJS) has produced a ratio By race. The adjustment ratios for white juvenile offending adjustment for major crime classes. BJS cautions, however, including adult co-offending are as follows: rape, ns; robbery, that this adjustment factor may not be suitable for adjusting ns; aggravated assault, 1.48; simple assault, 1.79; total vio- trends for subpopulations (such as juveniles) or specific esti- lence, 1.59. The adjustment ratios for white juvenile offending mates other than victimization rates. Consequently, adjust- excluding adult co-offending are as follows: rape, ns; robbery, ment factors for offender trends have been specifically de- ns; aggravated assault, 1.90; simple assault, 1.68; total vio- veloped for the analysis presented in this Bulletin. lence, 1.53. The adjustment ratios for black juvenile offending including adult co-offending are as follows: rape, ns; robbery, These adjustment factors were produced using the same ns; aggravated assault, 0.63; simple assault, 1.99; total vio- methodology employed by Lynch and Cantor (1996) to de- lence, 1.11. The adjustment ratios for black juvenile offending termine the adjustments BJS used for gross victimization excluding adult co-offending are as follows: rape, ns; robbery, trends. Specifically, offending rates were estimated using the ns; aggravated assault, 0.59; simple assault, 2.06; total vio- NCS and the NCVS data from the 1992 overlap. Signifi- lence, 1.53. The adjustment ratios for other-race juvenile of- cance tests were performed on the differences between the fending including adult co-offending are as follows: rape, ns; rate estimates from the old and new designs. When signifi- robbery, ns; aggravated assault, 2.29; simple assault, 3.75; cant differences were found, further tests were employed for total violence, 2.26. The adjustment ratios for other-race ju- specific subpopulations. If, for example, the rates from the venile offending excluding adult co-offending are as follows: NCVS were significantly different from the NCS for both rape, ns; robbery, ns; aggravated assault, 1.0; simple assault, males and females, the significance of this difference was 3.55; total violence, 3.18. tested to determine whether specific ratio adjustments should be applied for each gender. Whenever significant Although this adjustment procedure is useful, it must be used differences between the designs were found, adjustment with caution. The 18-month overlap period does not provide ratios were produced and applied to each subgroup. If these many data for estimating adjustment ratios. The NCVS sam- differences were not significant, then the same adjustment ple was not expanded during this period. Half the usual ratio was applied to all subgroups. The following are the ad- NCVS sample completed the pre-1992 instrument and half justment ratios used in this Bulletin for offending rates over- the post-1992 instrument. As a result, the reliability of the es- all, by gender, and by race. timates from the overlap period is less than that of the NCVS generally. This means that the adjustment ratios are likely to Overall. The adjustment ratios for overall juvenile offending be unstable for smaller population subgroups and rare offens- including adult co-offending are as follows: rape, not signifi- es. Moreover, the accuracy of the adjustment ratios is likely to cant (ns); robbery, ns; aggravated assault, ns; simple as- decrease the farther back in time one goes from the overlap sault, 2.02; total violence, 1.58. The adjustment ratios for period. For these reasons, this Bulletin presents trends only juvenile offending excluding adult co-offending are as fol- back to 1980, and these trends are limited to large classes lows: rape, 1.65; robbery, ns; aggravated assault, 0.91; of violent crime and major population groups. More detailed simple assault, 1.74; total violence, 1.29. trends were produced for the post-1992 period (see “Trends in Offending Rates Since 1993” on pages 8Ð11), when a consistent design was used for the full NCVS sample.

The trends in serious violent offending for the overrepresentation of male offenders some questions. It is tempting to infer that juveniles are broadly similar in the two and the overrepresentation of black of- police policy is the likely cause of these dif- data sources. Both show decreases in seri- fenders decreased. ferences, specifically, that the police tend ous violent offending until about 1986–87 to arrest black offenders more often than This examination of offending using NCVS and increases thereafter until around 1994, white offenders or female offenders more data confirms the general picture of of- when sharp decreases began and contin- often than male offenders. However, recent fending derived from UCR arrest data. At ued until 1998. Both race and gender dis- research (Pope and Snyder, in press) finds the same time, the greater racial dispro- proportionality declined between 1980 and no support for these contentions, at least portionality and the lower gender dispro- 1998 in both the UCR and NCVS data, as as they apply to serious violent crimes. portionality found in the arrest data raise

16 Other explanations must be considered. drop in juvenile co-offending was so steep adults), not from victims’ misidentifica- One factor may be differential reporting of for blacks compared with whites.9 Possible tion of offenders’ age. crimes by victims. For example, if victims explanations for this dramatic decrease in 5. All comparisons of statistics made in of crime by black offenders are more likely violent co-offending among black juveniles this report are statistically significant at to report their victimization to the police, should be explored empirically to en- the 0.1 level or less. The standard errors then the arrest rate for black offenders will hance the understanding of the recent used in these statistical tests take into be higher even if the police treat every drop in overall crime, with particular at- account the complex sampling used in case about equally. Other factors may be tention to the disparate declines in lone the NCVS. differences in the scope, definitions, and and group offending. procedures used in police record systems 6. This pattern is broadly consistent with and victim surveys. The UCR arrest rates, the theories of Blumstein and Wallman for example, include arrests for events that Endnotes (2000) concerning the increase in violence are not asked about in the NCVS, such as 1. The numerator for the arrest rates pre- in cities during the late 1980s and early commercial robbery, which in the UCR sented in this Bulletin is arrested juveniles 1990s. They contend that the increases arrest data is not separated from noncom- between the ages of 10 and 17. The de- in violence were due initially to the drug mercial robbery and thus cannot be re- nominator for the arrest rates is all juve- trade. Teenagers were recruited to par- moved from the dataset to allow equal niles between ages 12 and 17. This denom- ticipate in drug distribution, presumably comparison with the NCVS data. If these inator was used to be consistent with the with adults. This would increase adult events involve black offenders more than denominator for the offending rates from co-offending. Participation in drug distri- white offenders, then this could account the NCVS. Because there are relatively few bution required weapons, and the wide- for the greater racial disproportionality in arrested persons younger than 12 in the spread carrying of weapons encouraged the arrest data relative to the offending UCR data, including them in the numera- even those not in the drug trade to arm data. Similarly, the fact that larger urban tor (but not the denominator) of the rate themselves. The availability of arms places (which have disproportionately will not distort the arrest rate by much. among teenagers led to increases in se- large black populations) report more reli- The UCR arrest rates used in this Bulletin rious violence among teenagers not in- ably than smaller places in the UCR may will be slightly higher than they should be. volved in drug distribution. This could explain increases in juvenile-only offend- lead to the overrepresentation of black ju- 2. For ease of reading and to maintain veniles in the UCR arrest data. If analysts ing after the initial increases in adult consistency within the Bulletin, rounded co-offending with juveniles. adjusted the UCR arrest data to take the numbers are used in calculations and pre- missing information into account more sented in the text. The methodology used 7. The other serious violent crime catego- systematically, the differences between to calculate the trends presented in the ry (forcible rape) is relatively rare (1.12 the police and victim survey data in race Bulletin is described in “Computing Of- per 1,000), and trends are unstable. and gender disproportionality might fending Rates With the NCVS” (see pages 8. The percentages for adult co-offending disappear. 4–5). See also “Adjusting NCVS Data” were calculated from peaks in 1994. (page 16). All of these explanations—and others— 9. As Blumstein and Wallman (2000) argue, must be pursued to understand the ap- 3. The Survey of Inmates in State and Fed- the routinization of the drug trade and the parently discrepant pictures of offenders eral Facilities includes information on the decrease in violence attendant to the drug emerging from the UCR and the NCVS. The crime event that brought the inmate to trade may account for this. If juveniles good news is that most of these alternative prison, but the respondents are largely were recruited into the drug trade, it is explanations for the observed differences adults. not unreasonable to assume that they par- can be examined with the UCR and NCVS ticipated in groups (even gangs) distribut- data. Although beyond the scope of this 4. The imprecision of this statement comes, in part, from the difficulty of judg- ing drugs. In addition, it is reasonable to analysis, it is possible, for example, to de- assume that they would also participate velop and apply adjustments in the UCR to ing the age of offenders in the course of a criminal event and the fact that most in the violence attendant to the drug account for the overrepresentation of large trade in groups. As the drug trade be- population places. Then this explanation adult co-offenders are young adults. Vic- tims may overestimate or underestimate came more organized and less violent, the for black overrepresentation in the UCR need for group violence diminished. Even data would be eliminated. the age of some people in the group of offenders. In some cases, they may per- if these juveniles did not participate in the Finally, this examination of victim data of- ceive adults to be juveniles; in others, drug trade, living in areas with high levels fers a view of co-offending that is not often they may perceive juveniles to be adults. of violence virtually requires affiliation obtained (or obtainable) from police data. There is no reason, however, to believe with a group for protection. This, in turn, The majority of juvenile violent offending that these errors in identifying the age of increases the chance of co-offending. is done in groups. About one-third of this offenders will be in one direction. Indeed, group offending involves adults. Male juve- it seems more likely that these errors will References niles are more disposed to group offending be in both directions and will therefore be Biderman, A.D., and Lynch, J.P. 1991. than female juveniles, and black juveniles offsetting to some degree. This would Understanding Crime Incidence Statistics: much more than white juveniles. The de- mean that any ambiguity in distinguishing Why the UCR Diverges From the NCS. New clines in the juvenile offending rates for “juvenile offending” comes largely from York, NY: Springer Verlag. serious violent crimes between 1994 and the prevalence of co-offending or group 1998 were greatest for juvenile co-offending, offending (specifically group offending Blumstein, A., and Wallman, J. 2000. The especially for blacks. It is unclear why the that involves both juveniles and young Crime Drop in America. New York, NY: Cambridge University.

17 Dodge, R. 1987. Series Crime: Report of a of Justice Programs, Bureau of Justice American University, School of Public Field Test. Technical Report. Washington, Statistics. Affairs. DC: U.S. Department of Justice, Office Race as of Justice Programs, Bureau of Justice Lynch, J.P., Berbaum, M., and Planty, M. Pope, C., and Snyder, H. In press. Investigating Repeat Victimization a Factor in Juvenile Arrests. Statistics. 1999. Bulletin. Wash- With the National Crime Victimization ington, DC: U.S. Department of Justice, Federal Bureau of Investigation. 1999. Survey. Report to the National Institute Office of Justice Programs, Office of Juve- Crime in the United States, 1998. Washing- of Justice. nile Justice and Delinquency Prevention. ton, DC: U.S. Department of Justice. Lynch, J.P., and Cantor, D. 1996. Models Rand, M., Lynch, J.P., and Cantor, D. 1997. Jarvis, J.P., and Lynch, J.P. 1996. Towards for adjusting the NCS trend to account for Long Term Trends in Crime Victimization. an error profile of the Uniform Crime design differences between the NCS and Washington, DC: U.S. Department of Jus- Reports: Assessing the effects of imputa- the NCVS. Memorandum to the American tice, Office of Justice Programs, Bureau tion. Unpublished paper presented at The Statistical Association Committee on Law of Justice Statistics. Workshop on the Quantitative Analysis of and Justice Statistics, May 28, 1996. Criminal Victimization Crime and Justice Data, Inter-university Rennison, C. 2000. Bridging the Gap in Police in the United States, 1999. Consortium for Political and Social Re- Maltz, M. 1999. Washington, DC: Crime Data. search, University of Michigan, August Washington, DC: U.S. Depart- U.S. Department of Justice, Office of Jus- 2, 1996. ment of Justice, Office of Justice Pro- tice Programs, Bureau of Justice Statistics. grams, Bureau of Justice Statistics. Kindermann, C., Lynch, J.P., and Cantor, D. Singer, S. 1981. Homogeneous victim- 1997. The Effects of the Redesign on Vic- Planty, M.G. 2000. Examining the differen- offender populations: A review and some Journal of Criminal timization Estimates. Data Brief. Washing- tial returns from robbery incidents. Un- research implications. Law and Criminology ton, DC: U.S. Department of Justice, Office published dissertation. Washington, DC: 72(2):779–788.

Related Online OJJDP Publications If you are interested in more information about juvenile vio- perpetrated against animals and people. Available at lence and victimization, please consult the recently complet- ojjdp.ncjrs.org/pubs/violvict.html#188677. ed OJJDP publications listed below. These publications are Gun Use by Male Juveniles: Research and Prevention available in Adobe Acrobat and HTML online formats and (NCJ 188992). This OJJDP Bulletin examines patterns of gun can be downloaded from OJJDP’s Web site (ojjdp.ncjrs.org). ownership and gun carrying among adolescents, drawing on Click on “Publications” to access an alphabetical list of titles data from OJJDP’s Rochester Youth Development Study. The or to browse documents by topic area. You may also order Bulletin, which is part of the Youth Development Series, also copies of these and other publications by calling the Juve- addresses the interrelationship between gangs and guns and nile Justice Clearinghouse at 800Ð638Ð8736. describes prevention programs, including the Boston Gun Addressing Youth Victimization (NCJ 186667). This Action Initiative, the Office of Community Oriented Policing Services’ Plan Bulletin explains that although violent crime has de- Youth Firearms Violence Initiative, and OJJDP’s Partnerships creased significantly since publication of Combating Vio- To Reduce Juvenile Gun Violence Program. Effective efforts lence and Delinquency: The National Juvenile Justice Action to reduce illegal gun carrying and gun violence among youth Plan in 1996, too many children continue to be victims of require the support and participation of multiple community violent crime, child abuse and neglect, and exposure to vio- agencies. The information presented in this Bulletin is intend- lence. In turn, this victimization perpetuates a cycle of vio- ed to enhance those efforts. Available at ojjdp.ncjrs.org/ lence in which children later become the perpetrators of pubs/violvict.html#188992. violence against others. This Bulletin describes a variety of Juvenile Delinquency and Serious Injury Victimization programs and services that have been implemented at the (NCJ 188676).This OJJDP Bulletin draws on data from two State and local levels to address youth victimization and its OJJDP longitudinal studies on the causes and correlates effects. Federal agencies are collaborating to support this of juvenile delinquency—the Denver Youth Survey and the work through efforts such as the Coordinating Council’s call Pittsburgh Youth Study—to explore the interrelationship be- to break the cycle of violence. Available at ojjdp.ncjrs.org/ tween delinquency and victimization. The Bulletin, part of pubs/violvict.html#186667. OJJDP’s Youth Development Series, focuses on victims of Animal Abuse and Youth Violence (NCJ 188677). This violence who sustained serious injuries as a result of the OJJDP Bulletin describes psychiatric, psychological, and victimization. Being victimized may lead to victimizing others. criminal research linking animal abuse to interpersonal vio- The studies found that many victims were prone to engage lence perpetrated by juveniles and adults. Particular atten- in illegal activities, associate with delinquent peers, victimize tion is focused on the prevalence of cruelty to animals by other delinquents, and avoid legal recourse in resolving con- children and adolescents and to the role of animal abuse flicts. A clearer understanding of the patterns and predictors as a possible symptom of conduct disorder. In addition, the of victimization offers the potential for increased effectiveness motivations and etiology underlying the maltreatment of in designing and implementing strategies to reduce both vic- animals are thoroughly reviewed. The Bulletin includes rec- timization and offending. Available at ojjdp.ncjrs.org/pubs/ ommendations to curb such cruelty, while providing contact violvict.html#188676. information for additional resources concerned with violence

18 Snyder, H. 2000. Juvenile Arrests 1999. Washington, DC: U.S. Department of Jus- Share With Your Colleagues tice, Office of Justice Programs, Office Unless otherwise noted, OJJDP publications are not copyright protected. We en- of Juvenile Justice and Delinquency courage you to reproduce this document, share it with your colleagues, and reprint Prevention. it in your newsletter or journal. However, if you reprint, please cite OJJDP and the Snyder, H., and Sickmund, M. 1999. Juve- authors of this Bulletin. We are also interested in your feedback, such as how you nile Offenders and Victims: 1999 National received a copy, how you intend to use the information, and how OJJDP materials Report. Washington, DC: U.S. Department meet your individual or agency needs. Please direct your comments and questions of Justice, Office of Justice Programs, Of- to: fice of Juvenile Justice and Delinquency Juvenile Justice Clearinghouse Prevention. Publication Reprint/Feedback Thornberry, T., and Krohn, M. 2000. The P. O. Box 6000 self-report method for measuring delin- Rockville, MD 20849Ð6000 quency and crime. In Criminal Justice 800Ð638Ð8736 2000: Measurement and Analysis of Crime 301Ð519Ð5600 (fax) and Justice, vol. 4, edited by D. Duffee, E-mail: [email protected] D. McDowall, B. Ostrom, R. Crutchfield, S. Mastrofski, and L. Mazerolle. Washing- ton, DC: U.S. Department of Justice, Office of Justice Programs, National Institute of Justice.

This paper was prepared with funds provided to the National Center for Juvenile Justice by the Office of Juvenile Justice and Delin- quency Prevention to support its National Juvenile Justice Data Analysis Project (1999–JN–FX–K002). Points of view or opinions expressed in this Want to know more about the issues document are those of the author and do not necessarily represent the official position or It’s Fast in this Bulletin or related information? policies of OJJDP or the U.S. Department of Log on to ojjdp.ncjrs.org: Justice. ➤ Browse titles alphabetically or The Office of Juvenile Justice and Delinquency Prevention is a component of the Office of It’s Easy by topic. Justice Programs, which also includes the Bureau of Justice Assistance, the Bureau of ➤ Discover the latest OJJDP releases. Justice Statistics, the National Institute of Justice, and the Office for Victims of Crime. ➤ Subscribe to OJJDP’s listserv It’s Free JUVJUST and the electronic Acknowledgment newsletter JUSTINFO.

James P. Lynch is a Professor in ➤ Link to the NCJRS Abstracts Data- the Department of Justice, Law, and Society at American University. base to search for publications of interest.

19 U.S. Department of Justice PRESORTED STANDARD Office of Justice Programs POSTAGE & FEES PAID Office of Juvenile Justice and Delinquency Prevention DOJ/OJJDP PERMIT NO. GÐ91

Washington, DC 20531 Official Business Penalty for Private Use $300

Bulletin NCJ 191052