Conduct disorder (CD) and oppositional defiant disorder (ODD) symptoms are common among children and adolescents (e.g., Boylan et al. 2007; Loeber et al. 2000; Maughan et al. 2004) and are associated with a wide range of mental health problems later in life, including anxiety, mood, and substance use disorders (e.g., Fergusson et al. 2005; Kim-Cohen et al. 2003; Nock et al. 2007). According to the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association 1994), CD refers to “a repetitive and persistent pattern of behavior in which the basic rights of others or major age-appropriate societal norms or rules are violated” (p. 85) and ODD reflects “a recurrent pattern of negativistic, defiant, disobedient, and hostile behavior toward authority figures” (p. 91). As noted by Angold et al. (1999) in their meta-analysis, older studies have often collapsed CD and ODD into a single diagnostic category. More recent research has largely shifted toward conceptualizing these disorders as distinct entities (e.g., Burns et al. 1997; Greene et al. 2002; Nock et al. 2007; Rowe et al. 2010). However, the extent to which risk profiles or correlates of CD and ODD symptoms overlap among youth continues to be debated (e.g., Boden et al. 2010; Burke et al. 2002). The purpose of this study was to expand this line of research by examining the extent to which various individual and psychosocial correlates of CD and ODD symptoms overlap, using data from a large, representative community sample of African-American, Latino, and White boys and girls.

Several conceptual views have been offered in regard to the similarity of correlates and risk factors of CD and ODD symptoms. At least three general strands of models can be distinguished. A first approach argues that CD and ODD are mostly a function of common risk factors, such as similar underlying genetic makeup (e.g., Eaves et al. 2000). A second, somewhat similar approach, referred to as the differential severity model by Rowe et al. (2002), is based on a linear gradation hypothesis. It posits that the risk factors for CD and ODD are similar (e.g., social disadvantage, parental psychopathology), but that CD is more strongly related to these risk factors than is ODD (e.g., Frick et al. 1992; Rey et al. 1988). A third approach hypothesizes that both common and unique influences are related to CD and ODD (e.g., Dick et al. 2005; Tuvblad et al. 2009; Waldman et al. 2001). Investigations of the utility of these conceptual approaches and their variants are ongoing.

Empirical studies have identified a broad range of individual and psychosocial correlates and risk factors that are related to CD and ODD symptoms among children and adolescents (for a detailed review of this literature, see Burke et al. 2002). Selected findings are summarized below, focusing mostly on domains relevant for the current study. Individual correlates and risk factors include male gender (e.g., Bird et al. 2001; Maughan et al. 2004); race/ethnicity, with African-American children exhibiting higher symptomatology (e.g., Bird et al. 2001; Fabrega et al. 1993; Lahey et al. 1995); temperamental factors, such as inflexibility, intense and reactive responding, and low inhibitory control (Frick and Morris 2004; Nigg 2006; Rettew et al. 2004; Sanson and Prior 1999); impulsivity and attention deficit/hyperactivity disorder (ADHD; e.g., Greene et al. 2002; Maughan et al. 2004); and poor social skills or impaired social functioning (e.g., Greene et al. 2002; Matthys et al. 1995). Psychosocial correlates and risk factors include living in adverse or impoverished socioeconomic conditions (e.g., Boden et al. 2010; Greene et al. 2002; Lahey et al. 1995; Loeber et al. 1995; Rowe et al. 2010); violence exposure at home and in the community (e.g., Boden et al. 2010; Ford et al. 1999; McCabe et al. 2005); and stressful life events (e.g., Bird et al. 2001; Tiet et al. 2001).

Although these individual and psychosocial correlates have been associated with both CD and ODD in several studies, results have not always been consistent. For example, Boden et al. (2010) used prospective data from a New Zealand birth cohort to examine whether various psychosocial risk factors, maternal smoking during pregnancy, male gender, and IQ were significant predictors of CD and ODD at ages 14–16. Although the risk factor profiles of CD and ODD were very similar, a noteworthy finding was that male gender was related only to CD. In addition, the strength of associations between individual predictors and CD and ODD differed significantly in a few cases (e.g., family socioeconomic disadvantage). Similarly, using longitudinal data from 1,420 individuals aged 9–21 years, Rowe et al. (2010) found that male gender was related to CD but not to ODD. These findings call into question results from some prior studies indicating that gender was a common correlate or risk factor of CD and ODD. Another illustrative example pertains to race/ethnicity, for which studies have shown mixed patterns of group differences (e.g., Bird et al. 2001; Fabrega et al. 1993; Lahey et al. 1995). These inconsistencies are likely a function of several study features, including the specific racial/ethnic subgroups represented in the sample, the particular racial/ethnic subgroup designated as the baseline category during data analysis, and the extent to which race/ethnicity is confounded with other sociodemographic factors (see, e.g., Lahey et al. 1995). Further clarification of these racial/ethnic group differences is needed.

Thus it remains unclear which of the three conceptual approaches described above offers the most useful framework for understanding the correlates and risk factors of CD and ODD symptoms. In general, there appears to be fairly robust empirical evidence for both commonality and specificity in correlates and risk factors of CD and ODD; in some studies common or shared correlates appear to be the predominant pattern (e.g., Rowe et al. 2002; 2010), whereas other studies point to a higher degree of specificity in the correlates of CD and ODD (e.g., Shanahan et al. 2008). Moreover, there is considerable uncertainty regarding which correlates or risk factors should be conceptualized as specific to either CD or ODD. As noted by Rowe et al. (2010), to date, the only clear exception appears to be child gender. In fact, they suggested that male gender may be a specific risk factor for CD, which is consistent with findings from some recent studies (e.g., Boden et al. 2010; Rowe et al. 2010). Some features of the extant empirical literature likely contributed to this situation. First, many findings come from clinically referred samples of boys. Additional research with community samples comprising both genders is important to evaluate whether these patterns of findings replicate to non-referred youths and to girls. Second, findings have been qualified in some studies by confounding factors such as comorbid ADHD and gender (see Burke et al. 2002). More systematic control of these confounding factors in empirical studies is needed to advance the extant research on common and specific correlates of CD and ODD. Third, few comparisons of the correlates of CD and ODD across different racial/ethnic subgroups have been reported in the literature. More systematic comparisons across racial/ethnic subgroups are necessary to determine whether correlate-outcome patterns replicate across different racial/ethnic subgroups. Fourth, some risk factor domains have been studied fairly extensively as potential correlates of CD and ODD (e.g., adverse parenting behaviors, social disadvantage), whereas others remain relatively understudied (e.g., temperament, social skills). More systematic inclusion of such constructs in research on common and specific correlates of CD and ODD is important to allow for a more balanced evaluation of the merits of the differing conceptual approaches described above. Some of these gaps in the literature were addressed in this study.

The purpose of this study was to examine the extent to which individual and psychosocial correlates of CD and ODD symptoms overlap in a large, representative community sample of African-American, Latino, and White fifth graders, both for the total sample and across racial/ethnic subgroups. All associations were controlled for gender and ADHD symptoms, which have been shown to be important confounding factors in prior research. Guided by the empirical literature summarized above, measures of household income as percentage of the federal poverty level (FPL), life events, and violence exposure were investigated as fairly well-established correlates of CD and ODD symptoms. Because findings from McCabe et al. (2005) point to the importance of considering violence exposure at more than one level of a child’s ecological context, measures of violence exposure at home and in the neighborhood were each included.

Specific dimensions of the child’s temperament assessing aspects of the child’s general activity level and self-regulation were examined as putative correlates of interest in a more exploratory fashion. Self-regulation includes effortful control of attention (e.g., persistence, nondistractibility) and flexibility as important subcomponents. Although high general activity level and low effortful control have been linked to elevated externalizing behavior problems, conduct problems, and delinquency (e.g., Frick and Morris 2004; Sanson et al. 2004; Windle 1992; Zhou et al. 2007), these temperamental dimensions are often not included in studies examining common and specific correlates of CD and ODD symptoms. Measures of social skills are also relatively scarce in this literature, even though poorer functioning in perspective-taking, empathy, moral reasoning, and social-information processing, as well as egocentric peer representations, have been linked to externalizing behavior, aggression, and conduct problems (e.g., Cohen and Strayer 1996; Dodge 2011; Matthys et al. 1995; Tavecchio et al. 1999), whereas better social competence and social skills, including higher levels of cooperation and self-control, have been inversely associated with externalizing behavior problems (Bornstein et al. 2010; Gresham et al. 2011). In view of these findings, measures of social skills also were included in this study as putative correlates of CD and ODD symptoms.

We are not aware of any prior studies that also included this entire set of correlates. Regarding the expected pattern of common versus specific correlates of CD and ODD symptoms, the available literature did not permit the formulation of firm hypotheses, with the exception of male gender, which was posited to be a specific correlate of child CD symptoms, after adjusting for the effects of all other correlates.

This study used data from a large, representative community sample of children from three racial/ethnic subgroups to examine the specificity and commonality in correlates of CD and ODD symptoms. Several main conclusions resulted from the findings. First, common correlates of CD and ODD symptoms emerged as the predominant pattern in this study, although there also was evidence for specificity in correlates. Regarding the latter, gender and household income were found to be linked only to CD symptoms after adjusting for other correlates. This finding was particularly noteworthy because it is consistent with results from other recent studies (e.g., Boden et al. 2010; Rowe et al. 2010), which indicate that male gender might be a specific risk factor for CD after various confounding factors have been taken into account. The current findings provide additional empirical support for these results from prior studies.

Several other variables emerged as common or shared correlates of CD and ODD symptoms, including ADHD symptoms, race/ethnicity, life events, responsibility, self-control, and cooperation, even after adjusting for the effects of all other correlates. The direction of the vast majority of these multivariate associations was consistent with their univariate associations and the extant literature (e.g., Bird et al. 2001; Bornstein et al. 2010; Greene et al. 2002; Gresham et al. 2011; Maughan et al. 2004; Tiet et al. 2001). Interestingly, the multivariate model findings indicated that ethnic-minority status, especially Latino race/ethnicity, was related to a lower rate of CD and ODD symptoms compared with White race/ethnicity. This fits well with research showing that ethno-cultural factors can influence thresholds for the acceptability of youth’s behavior and parents’ interpretation of youth’s mental health symptoms (e.g., Roberts et al. 2005; Weisz et al. 1997). For example, cultural values such as respeto and simpatía (e.g., Calzada et al. 2010; Triandis et al. 1984) might function as protective factors that result in lower rates of CD and ODD symptoms, especially among Latino children. However, it is important to note that some studies (e.g., Bird et al. 2001) have revealed other patterns of race/ethnicity differences in rates of CD and ODD symptoms. Such inconsistencies across studies might be the result of adopting different analytical approaches. In this study, DSM-IV symptoms were assessed using a dimensional approach, whereas Bird et al. (2001) examined diagnoses of DSM-III-R disorders. It is widely recognized that dimensional and categorical approaches are both valuable but have different advantages and disadvantages. For example, the creation of binary diagnosis status measures from continuous measures is often based on arbitrary cut-points and results in loss of statistical power (see Helzer et al. 2006; Moffitt et al. 2008). This can contribute to discrepancies in findings across studies that are using different analytical approaches (Fergusson and Horwood 1995; Rowe et al. 2010). Clearly, the question of whether CD and ODD symptoms vary by race/ethnicity awaits further clarification.

The significant positive multivariate association of responsibility with CD and ODD symptoms should be regarded as tentative because it only became significant when confounding factors were taken into account and because prior research would have predicted an inverse relationship. Post hoc analyses (not shown) were conducted to explore whether it is more appropriately encapsulated as a curvilinear relationship. They revealed that the quadratic term of responsibility was significantly negatively associated with CD and ODD symptoms in the total sample, after controlling for its linear effect and all other correlates. However, this effect was not replicated in any racial/ethnic subgroup and appeared to be based on relatively few children. Thus, further research is needed to better understand the nature of the relationship between responsibility and CD and ODD symptoms.

Second, with some important exceptions (e.g., cooperation, responsibility, life events, household income as percent of the FPL), the pattern of common and specific correlates of CD and ODD symptoms was replicated in a quite consistent manner across the three racial/ethnic subgroups that were included in this study. The degree of replication across racial/ethnic subgroups was particularly high for the correlates of ODD symptoms. These findings are even more remarkable considering that the White children in this study generally came from less disadvantaged backgrounds than did the Latino and African-American children. Although it is premature to draw firm conclusions because studies on this issue are scarce and because it is possible that CD and ODD correlates may differ among distinctive Latino subgroups (e.g., Island Puerto Ricans; see Bird et al. 2001), the findings from this study suggest that differential exposure to adverse background factors among racial/ethnic subgroups may not always translate into distinctive correlate–outcome patterns as a function of race/ethnicity. This is encouraging from an applied perspective because substantial overlap in correlates of CD and ODD symptoms across different racial/ethnic subgroups implies that it may be sufficient to target the same components in prevention and intervention programs, irrespective of race/ethnicity. However, it is important to keep in mind that the extensive overlap in correlates of CD and ODD symptoms across racial/ethnic subgroups might be a direct function of the predictor set that was examined in this study. Generality in correlates of CD and ODD symptoms across racial/ethnic subgroups, if further substantiated in independent research, also does not preclude the possibility that culturally-sensitive implementation is important for maximizing intervention effectiveness in distinct racial/ethnic subgroups. Results from the moderator-effects analyses might be used to identify the factors that could be targeted for maximizing intervention effectiveness in specific racial/ethnic subgroups (e.g., for Latino children, cooperative behavior could be targeted).

Third, measures of specific dimensions of the child’s temperament and social skills are relatively scarce in the extant literature and thus were examined as putative correlates of CD and ODD symptoms. The findings of this study showed that social skills were considerably more salient putative correlates of CD and ODD symptoms than were temperamental dimensions, which were largely not significantly related to either outcome when the effects of all other correlates were controlled. This suggests that measures of social skills should be more systematically included in future studies on the correlates of CD and ODD symptoms, in particular those capturing levels of self-control and cooperation of the child, which have also been found to be related to broad externalizing behavior problem constructs (Bornstein et al. 2010; Gresham et al. 2011). Observed associations for assertion, however, were generally nonsignificant and those for responsibility should be interpreted with caution for the reasons described above. Because the univariate associations of the temperamental dimensions with CD and ODD symptoms were significant, the lack of significant multivariate associations in this study does not necessarily imply their etiological irrelevance. Several theoretical conceptualizations posit that temperament may translate into externalizing behavior problems in an indirect fashion, including through applying maladaptive social skills and competencies during transactions with the social environment (e.g., Bates et al. 1991). Although rigorous tests of specific hypothesized pathways are needed, initial evidence is promising insofar as temperamental dimensions, such as high levels of effortful control, have been linked with adaptive social competence (see Zhou et al. 2012), and better social skills have been found to be inversely related to externalizing behavior problems (Bornstein et al. 2010; Gresham et al. 2011). Therefore, a plausible alternative interpretation of the role of temperamental and social skills dimensions in this study purports that distal temperament correlates became non-significant after proximal social skills correlates of CD and ODD symptoms were taken into account. This speculation should be explored in prospective analyses.

Fourth, in addition to one of the social skills dimensions (assertion), violence exposure at home and in the neighbor-hood were not related to CD and ODD symptoms when the effects of all other covariates were taken into account. It is noteworthy that the univariate associations of both violence exposure measures with CD and ODD symptoms were significant and consistent with those reported in the literature (e.g., Boden et al. 2010; Ford et al. 1999; McCabe et al. 2005), which supports their inclusion in this study. However, other correlates and socioeconomic background factors turned out to be more salient correlates in the multivariate analyses, demonstrating again the importance of controlling for confounding and/or more distal factors in this line of research.

The current findings should be viewed in the context of various study limitations. First, CD and ODD symptoms were assessed via parent rating scales. Information from independent diagnostic data was not available for this study, but would be useful to address concerns about informant bias and shared measurement variance. However, it should be noted that some correlates in this study were based on child (rather than parent) reports, which alleviates these concerns to some extent. Second, the study was conducted with a large, representative community sample of African-American, Latino, and White fifth graders attending public schools in three metropolitan areas in the United States. Thus, study findings may not readily generalize to clinical samples, as well as to other racial/ethnic subgroups (e.g., Asian-Americans, Native Americans), age groups, and geographical areas. Similar research with these underrepresented groups is important for cross-validation purposes. Third, the conceptual overlap between some temperamental dimensions and psychopathology symptoms is problematic because it might have inflated observed associations between the two constructs (e.g., Nigg 2006; Rettew et al. 2004; Sanson and Prior 1999; Sanson et al. 2004). However, it is unlikely that this greatly affected the main conclusions from this study because temperament–psychopathology associations usually change little after the removal of overlapping items (e.g., Lemery et al. 2002). Fourth, the reliability of some measures (e.g., CD symptoms) used in this study was relatively low. However, this is not unexpected for measures that assess symptoms with low base rates (such as cruelty towards animals which is a symptom of CD). Finally, the cross-sectional nature of the study design does not permit inferences about the directionality of the observed associations. Longitudinal research is needed to further substantiate the direction of the associations between individual correlates and CD and ODD symptoms and to elucidate the underlying processes that account for such associations.

Despite these limitations, this study addressed important unresolved questions in the literature on common and specific correlates of CD and ODD symptoms. Future research on this issue should more systematically include measures of social skills and examine other dimensions of temperament and self-regulation that were not available in this study (e.g., deficits in executive functioning, emotional dysregulation, negative reactivity/emotionality). Comparison of common and specific correlates of CD and ODD symptoms by gender is also warranted.