Longitudinal Examination of the Relation Between Co­Offending with Violent Accomplices and Violent Crime by Kevin P. Conway and Joan McCord
(2001), is available at www.ncjrs.org/pdffiles1/nij/grants/192289.pdf.

Juveniles who commit crimes typically commit them in the company of their peers. This basic fact has been regularly reported in the literature since the late 1920s. Nevertheless, with rare exceptions, contempo­rary research focuses almost exclusively on juvenile delin­quents as individual actors.Indeed, police records tend to undercount co­offending, and published crime rates rarely take co­offending into account.Most crime rates are com­puted from individuals, with an assumption that each criminal event reported by or about an individual repre­sents a crime event (see “Measuring Juvenile Crime”).

Yet co­offenders provide a basis for multiple reports of single crime events. Not only are those who first offended before age 13 most likely to be co­offenders, but also the sizes of their offending groups (from 2 to 30 in thecurrent study) tend to further exaggerate the contributionsof youthful offenders to crimes. This exaggeration seems to contribute to a fear of youths that may be counterproductive.

Analyses that consider both co­offending and age at first arrest show that youthful offenders are most at risk for subsequent crimes if they commit their crimes with accomplices. Although very young offenders are responsible for a high pro­portion of juvenile crimes, their annual crime rate isnot particularly high unless they are co­offenders. Violence appears to belearned in the company of others. Those who commit crimes with violent offender even if the group does not commit violent crimes, are likely to subsequently com­mit violent crimes. This sug­gests that young offenders pick up attitudes and values from their companions.

To address issues raised by co­offending, including whether co­offending increases violence, the National Institute of Justice sponsored a study in Philadelphia that examined the criminal histories of a random sample of juvenile offenders. This Research in Brief discusses the study's findings and implications, considering four questions:

Why consider co­offending?
How is co­offending related to the age of offenders? 
How is co­offending related to recidivism?
How is co­offending related to violence?

Why consider co­offending?

Co­offending distorts report­ed crime rates by equating number of offenders with number of incidents and may increase a juvenile's risk for committing violent crimes through association with vio­lent peers. Statistics on crimes typically are based on the number of criminals  accused or convicted of crimes. Even when self­ reports are used, they indi­cate only which individuals within a stipulated population have committed crimes. Such statistics create a distorted picture of crime because many crimes are committed by more than one criminal and the proportion differs among different groups.

The distortion can be seen in the rare instances when crimes by lone offenders have been separated from those committed by multiple offenders. For example, the Sourcebook of Criminal Jus­tice Statistics, 2001, reports that 64 percent of the violent crimes attributed to lone offenders were committed by white offenders, but only 51 percent of the violent crimes attributed to multiple offenders were committed by offenders in “all white” 3 groups. These figures sug­gest that nonwhites are more likely to offend in groups. Therefore, crime rates based on arrests may exaggerate the contributions of non­whites to crime in the United States. The distortion has a particu­larly strong effect for juvenile crimes. In 1997, for example, the Supplemental Homicide Reports indicated that 44 per­cent of murders known to involve juveniles involved more than one perpetrator.

According to the Bureau of Justice Statistics, 23 percent of violent crimes in 1999 attributed to lone offenders were committed by juveniles under the age of 18, whereas over 40 percent of violent crimes attributed to multiple offenders were committed by juveniles. The fact that particular crimes are committed by more than one criminal not only distorts the connection between criminals and crimes, but also distorts estimates of effects from various crime preven­tion policies.

For example, researchers questioning the focus on incapacitation of high ­rate offenders noted that offenders' crime rates would be exaggerated if they had committed a large proportion of their crimes in groups. To more accurately measure the effect of incapacitation on crime rates, attention also must be given to the contin­ued criminal involvement of the co­offenders who remain in the community. In addition to distorting crimerates based on individuals and distorting the effects of intervention policies, co­offending may actually increase participation in crimes. Furthermore, the present study provides evi­dence that co­offending may increase violence (see “How co­offending is related to violence,”).

How co­offending is related to the age of offenders

Because prior evidence sug­gests that youths who start offending early commit more crimes than those who start late, effects of the age of firstcriminality should be consid­ered along with co­offending.

Most offenders in the Phila­delphia study committed their first offense between the ages of 13 and 15. Researchers identified youthswho committed a crime before the age of 13 as “young starters” and those who committed a first crime after age 15 as “late starters. They noted a relative decline in co­offending in relation to age, but this reflects a sharp increase in the number of crimes committed by single offenders rather than a decline in the number of co­offenses (see exhibit 1).

From ages 10 to 17, crimes committed alone, in pairs, and in groups increased. The number of crimes committed alone increased more rapidly than the number of crimes committed with accomplices. Rates for pairs and for groups were almost identical after age 14. When researchers differenti­ated property crimes from 9 violent crimes, they found a decline in co­offending after the age of 15 for property offenses (see exhibit 2). This decline, however, was paral­leled by a rise in solo proper­ty offending. Co­offending violence increased through­out adolescence, while solo violent offending leveled off around age 15. Among 16­and 17­ year­ old offenders, violent crimes were almost twice as likely to be co­offenses as solo offenses.

The youngest offenders at first arrest were the most likely to mix co­offending and solo offending, but least likely to commit all their crimes alone. Those first arrested at ages 16 or 17, on the other hand, were most likely to commit crimes alone. About 40 percent of offenders com­mitted most of their crimes with accomplices, regardless of their age at first arrest (see exhibit 3).

How co­offending is related to recidivism

The Philadelphia delinquents first arrested when they were under 13 years of age had higher rates of recidivism than those first arrested when they were older. Co­ offending, however, distorts the picture of recidivism because there are actually fewer crime incidents than individual crime rates indicate (see exhibit 4). Specifically, crime rates are inflated if co­offending is not taken into
account. In contrast, crime rates that account for co­offenders count each crime incident once even if multiple offenders have been arrested for the crime.

The crime ­incident ratio, which accounts for co­offending, is greatest for the young starters— indicating that crime rates for young delinquents are most likely to be inflated when co­offending is not taken into account. Study findings on recidivism provide a good example of the increased information that comes from recognizing co­offending. The number of Index crimes was consistent­ly higher for delinquents who co­offended at least 25 per­cent of the time. This pattern was particularly evident for the young starters. The young starters who co­offended at least 25 percent of the time were arrested for almosttwice as many Index crimes as the young starters who typically committed solo 10 offenses.

Thus, the number of arrests for Index crimes reflected both the age at first arrest and the proportion of crimes that were co­offenses (see exhibit 5), revealing that young­starter delinquents who mostly co­offend com­mitted the most crimes. An examination of annualcrime rates further demon­strates how crime rates can be inflated by inattention to co­offending. In each category of age for first arrest, individ­ual co­offending rates were higher than solo rates (see exhibit 6). The offenders first arrested at ages 16 and 17 had the highest rates for both solo and co­offenses. However, these high recidivism rates are due to both the compressed duration of their measured criminal activities and the fact that such a high proportion of their crimes are co­offenses.

Despite committing crimes at lower rates, the offend­ers who had first been arrested under the age of 13 had the highest ratio of co­offending to solo offend­ing. But young starters arenot high recidivists if one considers the length of time
they are exposed to the 11 juvenile justice system.  These analyses show not onlythat crime rates based on indi­viduals are most inflated for  young­starting delinquents, but also that targeting youth­ful co­offenders could be the most productive approach to reducing future crime.

How co­offending is related to violence

Those who generally committed crimes with others were more likely to commit violent crimes than were solo offenders. The association between co­offending and violence was strongest for young starters. Young starters. On average, offenders who had accom­plices for at least 25 percent of their crimes and had been arrested before the age of 13 committed more than two violent crimes (see exhibit 7).  Young starters who commit­ted most of their crimes alone, however, were not particularly prone to committing violent crimes.

On the other hand, co­offending young starters were considerably more likely to commit violentcrimes than were late starters, specially late starters who mostly worked solo. Thus, because the vast majority of young starters commit many of their crimes with others, the effects of age and co­offending on vio­lence tend to be confounded.

Is violence learned?

The association between co­offending and violence raises the question of whether kids who tend to be violent hang out together and therefore commit violent crimes or whether learning accounts for some of the high level of violence. To test the latter, researchers identified 236 offenders in the random sample of 400 who had not committed violent crimes before committing a crime
with others.

These offenders committed their first co­offenses with 514 accomplices. Groups ranged from 2 to 15 offend­ers. Pairs committed 42 percent of these crimes. Co­offenders typically matched their accomplices in ethnic 12 identity. Age comparisons revealed that most of the offenders identified in their first co­offense were younger 13 than their accomplices. Among the 236 offenders who had not been violent before their first co­offense, 90 participated in a violent first co­offense; among these, 62 percent committed at least one additional violent offense after this first one.

Another 61 juveniles partici­ pated in a nonviolent co­offense with co­offenders who had previously been violent. These juveniles were even more likely to subsequently commit a violent crime than those who had actually participated in a violent crime for their first 14 of offense. To check whether peer conta­15 gion may have influenced the learning of violence, researchers divided the previ­ously nonviolent offenders who committed a first co­ offense that was not violent into two groups according to whether their accomplices had been violent before the target co­offense.

Those who committed a nonviolent offense with violent people were considerably more likely to commit a subsequent vio­ lent crime—80 percent of those with violent accom­plices, compared with 56 percent of those with only nonviolent accomplices, com­mitted at least one violent 16 crime after the co­offense.The data showed no system­atic relationship between age at first offense and whether or not nonviolent offenders co­offended with violent offenders for a first co­offense. Nevertheless, both whether a violent offender participated in the first co­offense and age at first arrest predicted whether a previ­ ously nonviolent offender would commit a violent crime (see exhibit 8).

Committing a first co­offense with violent accomplices contributed to the likelihood that violent crimes would be committed, regardless of age at first arrest. That is, violent peers increase the likelihood that nonviolent offenders will commit violent offenses. How may violence be learned? Peer delinquency seems to be more than a training process for learn­ing how to be delinquent. Interaction among delin­quent peers apparently encourages and escalates their proclivity to commit crimes. Co­offenders may learn through the influence of violent accomplices that violence can be an effective means for getting money or satisfying other desires.

They also may learn that insults or fear provide adequate grounds for violence. An adequate theory of crime should take into account both how others influence ndividual behavior and how individuals selectively seek companions who are likely to promote criminal behavior. Construct Theory postulates that co­offending provides a young offender justification for continued delinquency, encouraging him or her to seek out accomplices and commit additional crimes(see “Construct Theory”). This implies that interven­tions need not be directed at deep­seated emotions. Rather, behavioral change can be expected as a conse­quence of changing beliefs in relation to grounds for action.

Implications for policy and practice Because many juvenilecrimes are committed in the company of others, crime rates cannot be accuratelyportrayed unless co­offendingis accurately recorded. Yet inspection of official recordsindicates that attention hasnot focused on this feature of crime events. Too often, acrime is considered to besolved when a single arresthas been made. The Philadelphia study demonstrates that crime records should contain accu­ rate information about co­offending.

Such accuracy is necessary if the effects of policy shifts are to be meas­ured or if differences in crime rates are to be used as a basis for such preventive actions as deploying police and implementing target­hardening measures. Perhaps the greatest chal­lenge for intervention is to target youthful co­offenders in a way that reduces the likelihood that they will devel­op attitudes that promote crime. The study's findings imply that lessons of violence are learned “on the street,” where knowledge is passed along through impromptu social contexts, including those in which offenders commit crimes together.

Interaction among delinquent peers apparently serves to instigate crimes and to esca­ late their severity. More research on this issue is warranted, especially stud­ies that measure peer influ­ ence on intentional action, track the selection of accom­plices across multiple crimes, examine the learning processes involved in the transfer of violence across offenders, and identify indi­vidual offenders who may be particularly susceptible to (or unaffected by) the influence of violent accomplices. When developing and evaluat­ing strategies designed to prevent or reduce violence, practitioners and evaluators may want to consider co­offending patterns, individuals' choices of accomplices, and factors that increase the risk of co­offending, especially among very young offenders.

Notes
1. A 1928 study found that 82 per­cent of juveniles brought to court in Cook County, Illinois, committed their offenses as members of groups. See Shaw, C.R., and H.D. McKay, Juvenile Delinquency and Urban Areas, revised edition, Chicago: University of Chicago Press, 1969 (first published 1942). For studies that focused on group processes to try to understand juve­ nile delinquency, see Cohen, A.K., Delinquent Boys, Glencoe: Free Press, 1955; Cloward, R.A., and L. Ohlin, Delinquency and Opportunity, New York: Free Press, 1960; and Short, J., and F.L. Strodtbeck, Group Process and Gang Delinquency, Chicago: University of Chicago Press, 1965.
2. See Carrington, P.J., “Group Crime in Canada,” Canadian Journal of Criminology (July 2002): 277–315; Hockstetler, A., “Opportunities and Decision: Interactional Dynamics in Robbery and Burglary Groups,” Criminology 39 (3) (2001): 737–763; McCarthy, B., J. Hagan, and L.E. Cohen, “Uncertainty, Cooperation and Crime: Understanding the Decision to Co­Offend,” Social Forces 77 (1) (1998): 155–184; Weerman, F.M., “Co­Offending as Social Exchange: Explaining Characteristics of Co­Offending,” The British Journal of Criminology 43 (2) (2003): 398–416.
3. Maguire, K., and Pastore, A.L., Sourcebook of Criminal Justice Statistics, 2001, Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics, 2003, NCJ 196438. Calculations have omitted “mixed” and “not known.”
4. Supplemental Homicide Reports are part of the FBI's Uniform Crime Reporting system. See also Snyder, H.N., and M. Sickmund, Juvenile Offenders and Victims: 1999 National Report, Washington, DC: U.S. Department of Justice, Office of Juvenile Justice and Delinquency Prevention, 1999.
5. Bureau of Justice Statistics, Sourcebook of Criminal Justice Statistics, 1998, Washington, DC: U.S. Department of Justice, 1999, NCJ 176356.
6. See Reiss, A.J., Jr., “Co­Offending and Criminal Careers,” in Crime and Justice, vol. 10, ed. N. Morris and M. Tonry, Chicago: University of Chicago Press, 1988: 117–170.
7. See Hindelang, M.J., “With a Little Help From Their Friends: Group Participation in ReportedDelinquency,” British Journal of Criminology 16 (1976): 109–125; and Reiss, A.J., Jr., and D.P. Farrington, “Advancing Knowledge About Co­Offending: Results From [a] Prospective Longitudinal Survey of London Males,” Journal of Criminal Law and Criminology 82 (1991): 360–395. Also, delinquents in co­ offending groups studied in Japan reported that they committed more crimes together than alone. See Suzuke, S., Y. Inokuchi, K. Watanabe, J. Kobayashi, S. Okela, and Y. Takahashi, “Study of Juvenile Co­offending,” Reports of the National Research Institute of Police Science 36 (1995): 2, 64.
8. Before attention was drawn to co­offending, high recidivism rates had been linked with offenders who were particularly young when they began to commit crimes. See McCord, J., and K.P. Conway, “Patterns of Juvenile Delinquency and Co­Offending,” in Crime and Social Organization, vol. 10 of Advances in Criminological Theory, ed. E. Waring and D. Weisburd, New Brunswick, NJ: Transaction Publishers, 2002:16.
9. Property crimes were burglary, vehicle theft, theft other than vehicle, arson, vandalism, criminal trespass, forgery or counterfeiting, embezzlement, fraud, and risking or causing a catastrophe. Violent crimes were murder, attempted murder, rape, robbery, aggravated assault, simple assault, terroristic threaten­ing, intimidating a witness, prowling, and cruelty to animals.
10. Index crimes are eight categories of serious crime collected by the FBI's Uniform Crime Reporting Program. Violent Index crimes are homicide, criminal sexual assault, robbery, and aggravated assault/ battery. Property Index crimes are burglary, theft, motor vehicle theft, and arson.
11. If all young criminals spend about 5 years actively committing crimes, only those arrested before their 13th birthdays would spend all their crimi­ nal years as juveniles. To compen­sate for this potential bias, individual crime rates were computed for both solo offenses and co­offenses, on the assumption that once a juvenile committed a crime, he or she would remain a delinquent until the age of 18. Whatever bias this computation introduced affected solo and co­offending rates alike.
12. The ethnic identity of co­offenders and accomplices matched for 96 per­ cent of black offenders, 83 percent of white offenders, and 83 percent of Hispanic or other offenders. Re­searchers traced the criminal histories of 396 of the accomplices, a success rate of 77 percent.
13. Sixty­three percent were younger, 19 percent were older, and 18 percent were the same age or very close. 2
14. X =5.626, p<.02. (1)
15. For discussion of this issue, see Dishion, T.J., J. McCord, and F. Poulin, “When Interventions Harm: Peer Groups and Problem Behavior,” American Psychologist 54 (9) (1999): 1–10. 2
16. X =9.065, p<.003. (1)
17. Case studies and self­report data converge to suggest that delinquent groups socialize their members in ways that encourage and value violence.
18. See “Construct Theory” sidebar; also see McCord, J., “Understanding Childhood and Subsequent Crime,” Aggressive Behavior 25 (1999): 241–253.

For full report, sidebars and graphs please go to http://www.ncjrs.org/pdffiles1/nij/210360.pdf