involvement in inmate economy) which could illustrate how interactional patterns or lifestyle influence inmate behavior. An important recent study (Jiang & Fisher-Giorlando, 2002, pp. 349–352) that was able to compare the three theoretical approaches found that the situational model was the best predictor of inmate misconduct directed against prison staff and intra-inmate violence. However, they also found that based on the χ2 change per variable in each model that the deprivation model was the most powerful predictor of misconduct.Similarly, penologists (e.g. Gaes et al., 2002; Gaes et al., 2003; Wooldredge, Griffin,& Pratt, 2001) are increasingly advocating the use of both individual and institutional variables coupled with multilevel data analysis to better specify models of inmate behavior. Indeed, on this note, Gaes et al. (2003, p. 530) issued a challenge to investigators of inmate behavior to simultaneously examine variables at both units of analysis that encompass all three theoretical
THE AFFECTS OF GANG VIOLENCE IN AMERICA 8perspectives. We agree that the future for studies of inmate behavior is this more integrated, methodologically sophisticated approach. However, this does not mean that theoretical questions that can be examined with local data are fully settled. Indeed, our central findings indicated that although street gang members and persons involved in security threat groups were important threats to prison order, they paled in comparison to chronic offenders, racial and ethnic minorities, and inmates with weak familial ties or social support. As suggested earlier, that inmates whom were always involved in gangs were negatively involved in prison violence was, we presume, a function of the department of corrections to identify and appropriately supervise those inmates who posed the greatest gang threat.In fact, our findings largely conflict with the recent study by Gaes et al. (2002, p. 381) who found that gang membership increased violence and almost all other forms of prison misconduct whether these behaviors were rule infractions or actual crimes. Moreover, their gang effects were robust controlling for measures of violent risk, history of violence, and other relevant background factors.What accounts for these discrepant findings? Gaes et al.’s (2002) study was in many ways vastly superior to the current effort in that the sample was the entire male population of the Bureau of Prisons and they controlled for 27 separate gangs. Comparatively, the current effort used proxies of risk based on street gang, prison gang and interaction gang involvement. Thus, it is possible that the robust gang effect does not fully emerge until the various types of gangs are disaggregated. The discordant findings could also reflect the behavioral differences between state prisoners and inmates in the federal system.
THE AFFECTS OF GANG VIOLENCE IN AMERICA 9Still, the importation model of inmate behavior should be considered a project under construction. The current findings lend credence to Allender and Marcell’s (2003) advisement