Exploring Sex Differences among Sentenced Juvenile Offenders in Australia

Abstract

Recent research applying latent class analysis (LCA) reveals considerable diversity in the self‐reported offending patterns of incarcerated females and suggests that a failure to recognize these patterns will hinder the ability to understand mechanisms that lead females to serious offending. Using data from a cohort of serious juvenile offenders in Queensland, Australia, this paper extends the earlier research by using LCA to assess sex differences in juvenile criminal offending. Results indicate that female offenders are not a homogenous group with respect to their offending patterns, that there is a degree of symmetry between male and female offenders, and that childhood experiences of maltreatment increase the likelihood of membership in the most serious offending group for both males and females. Implications for theory and policy as well as directions for future research are highlighted.

Keywords:

• juvenile offenders
• sex differences
• child maltreatment
• Australian
• child protection
• latent class analysis

Acknowledgments

Our thanks to Cassia Spohn and to referees from Justice Quarterly for comments on an earlier draft. An earlier version of this paper was presented at the Stockholm Criminology Symposium in Stockholm, Sweden, in June 2009.

Notes

1. Our research uses official records of offending—i.e., proven matters entailing more filtering through the justice system—which offers a more constrained view of the true nature of criminal offending, especially when compared with self‐reports. Yet, both sources of offending are vulnerable to error (Thornberry & Krohn, 2003). For example, while official records capture official justice reactions, self‐reports are prone to over‐ and under‐reporting and tap into a wide array of trivial offenses. Further complicating matters is the reality that very few longitudinal criminal career studies contain both self‐reports and official records (Piquero et al., 2003). In a rare example, Brame, Fagan, Piquero, Schubert, and Steinberg (2004) examined the relationships between official and self‐report offending measures using two samples of serious adolescent offenders in Phoenix and Philadelphia. Their analysis indicated that (1) offense frequencies—whether measured by official or self‐report records—were positively skewed, (2) adolescents who were arrested more often tended to self‐report more offending than those who had been arrested less often, and (3) the correlation between self‐reported offending and official arrest frequency was robust across different demographic groups within the two samples. This led the authors to conclude that “In light of the complementary strengths and weaknesses of official record and self‐reported data sources, most contemporary researchers take it as axiomatic that both are important in advancing knowledge about criminal behavior. Thus, self‐report and official record data sources should not be viewed as competing vehicles by which to assess involvement in criminal offending. Rather, both methodologies tap into the same underlying behavior—they are just different indicators of that behavior” (2004, p. 258). Indeed, Farrington (1989, p. 418) noted that official records and self‐reports actually produce convergent inferences “on such important topics as prevalence, continuity, versatility, and specialization in different types of offenses”. We return to this issue later in the discussion section.

2. While our study employs an official measure of maltreatment, Odgers et al.’s (2007) study employed both official and self‐report measures of maltreatment. Of course, both self‐report and official measures have their strengths and weaknesses. As summarized by Smith, Ireland, Thornberry, and Elwyn (2008, p.174), two major advantages of official records include: (1) a reduction in the possibility of false‐positives because formal investigations rely on relatively objective criteria to assess harm to the child, and (2) the possibility that relatively trivial events will be filtered out from being classified as maltreatment. At the same time, official measures are also limited: (1) by the possibility of false‐negatives (i.e., classifying maltreated children and adolescents as ‘never maltreated’), and (2) by under‐representing the amount of actual maltreatment in the population because of the filtering that occurs in the reporting and determination of substantiated maltreatment. On the other hand, self‐report measures of maltreatment are designed to offset concerns associated with official records, as noted previously. For example, self‐reports are not subject to filtering and may capture a larger inventory of maltreatment experiences. Also, self‐report measures may be more practical and convenient compared with official measures. Yet, self‐report measures are also limited. For example, retrospective‐based measures ask adults about experiences during childhood and adolescence, thus raising issues about validity and reliability particularly when long recall periods are required (Widom, Raphael, & DuMont, 2004). Also, self‐report measures are potentially influenced by mood and psychopathology during assessment periods. In an important but rare study comparing official and self‐report maltreatment measures, Smith et al. (2008) used longitudinal data from the Rochester Youth Development Study (New York) to address the prevalence and concordance of each type of maltreatment measure, their relationship to social disadvantage, and their prediction to four antisocial outcomes in adolescence and early adulthood, including arrest, self‐reported violence, general offending, and illegal drug use. Several findings emerged from their study. First, self‐reported retrospective maltreatment was more prevalent (29% vs. 21%). Second, among those with official reports in young adulthood, about half self‐reported maltreatment, whereas 37% of those self‐reporting had an official report. Overall, these authors found that although both measures performed similarly in predicting early adult antisocial behavior, there were situations where one method would be preferable to the other. For example, if one wanted to know whether maltreatment experienced in childhood caused adolescent problem behaviors, the measure of maltreatment would need to precede estimates of problem behavior (Ireland et al., 2002). Thus, official maltreatment measure, recognizing its limitations, is more useful than retrospective self‐report measure to establish temporal linkages and causal order, as well as mediators and moderators, between maltreatment and outcomes of interest.

3. Since latent variable modeling does not rely on common assumptions that are easily violated, such as normality and linearity, it has advantages over related techniques, such as k‐means or hierarchical cluster analysis (see Vaughn, DeLisi, Beaver, & Howard, 2009).

4. We would like to thank an anonymous reviewer for suggesting that we comment more fully on how gender‐based responses to abuse may explain differences in class membership probabilities.

5. On this last point, an anonymous reviewer noted that it is possible that family therapy may affect later behavior differently than being placed in a foster home. The knowledge base on this issue remains controversial and unresolved (Donnelly & Daro, 2002) and requires further research.