Forging a Link Between Conduct Disorder and Adolescent/Adult Offending Via Externalizing Behavior and Reading Performance

Abstract

The current study tested whether teacher-rated externalizing behavior and academic (reading) performance mediate the relationship between childhood onset conduct disorder and self-reported adolescent delinquency and officially recorded adult offending. All 411 boys from the Cambridge Study of Delinquent Development served as participants in this study. Mediation analysis revealed that the direct effect of childhood onset conduct disorder and the indirect effect of reading performance on mid-adolescent delinquency and early adult offending were nonsignificant, but that adolescent externalizing behavior had a significant indirect effect on both delinquency and adult offending.

Keywords

chaining conduct disorder externalizing behavior reading performance delinquency

Over the past 20 years, epidemiological surveys have identified a high rate of conduct disorder in children and adolescents. The New York State Children in the Community Study, for instance, placed the 12-month prevalence of conduct disorder for boys aged 10–16 years at 16%, for girls aged 10–13 years at 4%, and for girls aged 14–16 years at 9% (Cohen et al., 1993). The lifetime prevalence of conduct disorder in 9- to 16-year-old boys and girls participating in the Great Smoky Mountains epidemiological study was 14% and 4%, respectively (Costello, Mustillo, Erkanli, Keeler, & Angold, 2003). According to the National Comorbidity Survey Replication, the lifestyle prevalence of conduct disorder is 12.0% in boys and 7.1% in girls (Nock, Kazdin, Hiripi, & Kessler, 2006). Prevalence rates for self-reported delinquency and official criminal offending are even higher than they are for childhood and adolescent conduct disorder. Self-reported rates of involvement in violent and property offenses for participants in the longitudinal Rochester, Denver, and Pittsburgh youth surveys were 15% at age 11 and nearly 50% at age 17 (Huizinga, Loeber, & Thornberry, 1993). High prevalence rates are not confined to self-report data, however. One third of all males and 14% of all females from the 1958 Philadelphia birth cohort study had an official arrest on file before age 18 (Tracy, Wolfgang, & Figlio, 1990).

One finding that came out of the original 1945 Philadelphia birth cohort study was that a minority of offenders are responsible for the majority of offenses (Wolfgang, Figlio, & Sellin, 1972). Building on this finding, Vaughn and colleagues studied several epidemiological samples and found evidence that about 5% of the population is responsible for the majority of problem behaviors exhibited by the overall cohort. Subjecting data from the 43,093-member National Epidemiological Survey on Alcohol and Related Conditions (NESARC) to latent class analysis, Vaughn and colleagues (2011b) identified a small group of adults (5.3% of the sample) who exhibited high levels of antisocial and externalizing behavior to include substance misuse. In a latent class analysis of youth data from the National Survey on Drug Use and Health (NSDUH), Vaughn, Salas-Wright, DeLisi, and Maynard (2014) determined that 4.7% of the sample were characterized by extreme externalizing behavior. These results coincide with Moffitt’s (1993) notion of an early onset pattern of antisocial and externalizing behavior that she refers to as life course persistent offending and with Walters’ (2014a) views on the relationship between drugs and crime in what he refers to as the “worst of both worlds” hypothesis. With respect to the latter, it has been found that individuals with histories of both substance misuse and antisocial behavior display significant worse future adjustment than individuals with histories of substance misuse alone or antisocial behavior alone (Walters, 2014a, 2015a, 2015b).

Studies assessing the relationship between childhood onset conduct disorder and later delinquency indicate that the relationship is strong. Given the presence of research showing that age of conduct disorder onset is a robust correlate of future offending (Murray & Farrington, 2010)—though not as stronger as the relationship between early police contact and future offending (DeLisi, Neppl, Lohman, Vaughn, & Shook, 2013)—it makes sense that childhood onset conduct disorder would more likely be followed by juvenile delinquency and adult crime than adolescent onset conduct disorder. Conduct disorder, regardless of whether the onset occurs in childhood or in adolescence, however, does not always lead to delinquency and crime. Whereas conduct disorder correlated with delinquency in a longitudinal study by Burke, Loeber, Mutchka, and Lahey (2002), the majority of boys who served as participants in this study either met the criteria for conduct disorder but not delinquency (53.2%) or met the criteria for delinquency but not conduct disorder (10.4%). Similarly, in a 30-year follow-up of former Norwegian child psychiatric inpatients, Mordre, Groholt, Kjelsberg, Sandstad, and Myhre (2011) determined that slightly less than half of the individuals who received a diagnosis of conduct disorder as children had an official arrest record as adults.

Therefore, while conduct disorder and delinquency are clearly related, they are far from identical. Waldman and Lahey (2010) list three reasons why conduct disorder and delinquency do not always coincide: (1) Not all juvenile offenses are symptoms of conduct disorder, (2) not all symptoms of conduct disorder are criminal offenses, and (3) conduct disorder requires the presence of at least 3 out of 15 antisocial symptoms within a 12-month period, whereas delinquency can be defined on the basis of a single criminal act. There is another possibility, one in which additional variables mediate the relationship between conduct disorder and later delinquency. Externalizing behavior, for example, has been found to correlate with both conduct disorder (Rolon-Arroyo, Arnold, & Harvey, 2014) and delinquency (Cohn, van Domburgh, Vermeiren, Geluk, & Dorelijers, 2012). In their general theory of crime, Gottfredson and Hirschi (1990) emphasize externalizing behavior in the form of low self-control and attribute it to weak parenting. Others have adopted a more biologically based view and consider externalizing behavior a manifestation of a heritable trait referred to as behavioral disinhibition (Krueger, Markon, Patrick, & Iacono, 2005; Young et al., 2009). Externalizing behavior is distinguished from conduct disorder by virtue of the fact that it encompasses several diagnoses and patterns other than conduct disorder (e.g., attention deficit/hyperactivity disorder, oppositional defiant disorder, antisocial behavior, and impulsive and sensation seeking personality traits). Externalizing behavior and conduct disorder are distinct from delinquency by virtue of the fact that their symptoms are not necessarily illegal.

Another possible mediator of the conduct disorder–delinquency relationship is academic or reading performance. There is a growing body of research to suggest that academic achievement, with specific reference to reading performance, correlates with both conduct disorder and later delinquency (Bennett, Brown, Boyle, Racine, & Offord, 2003; Maughan, Pickles, Haggell, Rutter, & Yule, 1996; Willcutt & Pennington, 2000; Williams & McGee, 1994). In addition, DeLisi, Vaughn, and their colleagues have uncovered a relationship between poor reading achievement and concurrent ratings and self-reports of psychopathic traits (DeLisi et al., 2011; Vaughn et al., 2011a). An important but unanswered question at this point is whether the relationship between reading performance and delinquency/psychopathy is causal. A recent series of longitudinal analyses found that prior general language ability correlated better with subsequent behavioral problems than prior behavioral problems correlated with subsequent general language ability (Petersen et al., 2013). This suggests that a causal connection may, in fact, exist between general language ability, to include language expression/mechanics and reading comprehension/recognition skills, and subsequent behavioral problems like attention deficit and externalizing behavior.

Previous research has shown that the relationship between temperament and early delinquency is mediated by externalizing behavior (Walters, 2014c). The purpose of the present study was to determine whether the relationship between conduct disorder and adolescent delinquency/adult offending is mediated by externalizing behavior and/or reading performance. This was accomplished using a five-wave longitudinal design in which the independent variable (childhood onset conduct disorder) was measured at the second wave, the mediators (externalizing behavior and reading performance) were measured at the next two waves, and the dependent variables (delinquency, adult offending) were measured in the final wave. It was hypothesized that while reading performance would correlate with proximal conduct disorder and externalizing behavior, it would be inferior to externalizing behavior in linking the independent variable (conduct disorder) to the dependent variables (adolescent delinquency and early adult criminality). Adopting the chaining model introduced in Walters (2014c) and using the contrast test proposed by Preacher and Hayes (2008), three comparisons were made. The first comparison involved contrasting the two-mediator externalizing chain with the two-mediator reading chain. The second comparison involved contrasting the first-stage mediator externalizing chain with the first-stage mediator reading chain. The third comparison involved contrasting the second-stage mediator externalizing chain with the second-stage mediator reading chain. In each case, it was hypothesized that the externalizing chain would achieve significance and would be significantly stronger than the reading chain in predicting both adolescent delinquency and early adult offending.

Method

Participants

Participants for this study were all 411 South London boys from the Cambridge Study of Delinquent Development (Farrington, 1995; West & Farrington, 1973). Most of the boys were born in 1953 and each was followed from age 8 through middle adulthood using interviews with the boys, their parents, their teachers, and their peers, accompanied by a review of official arrest/conviction records. This was not a probability sample but a census of 8- to 9-year-old boys from six state primary schools and one school for subnormal boys within a mile of the research office. The majority of boys were white in appearance and of British origin (87%). Remaining participants were either Afro-Caribbean or other European (German, Irish, Swedish, Australian, French, Spanish, or Portuguese). Over 90% of the boys came from intact working class homes in South London.

Measures

Independent variable

The independent variable for this study was a four-level rating based on an interview held with the child’s parents when the child was 10 or 11 years of age (Wave 2). This rating assessed the child for conduct disorder symptomatology using the following scale (1 = none; 2 = minimal; 3 = moderate; and 4 = severe). A minimal rating indicated that the child displayed minor problems limited to a single area of functioning (school, home, or court), which did not significantly impact on the boy’s relationships with others. A moderate rating reflected greater persistence of conduct disorder symptomatology, affecting several areas of functioning and causing conflict at home or in school. A severe rating denoted serious and repeated maladjustment in the boy’s behavior and relationships across a wide variety of settings and circumstances.

Mediating variables

The two mediating variables included in this study were teacher-rated externalizing behavior and reading performance during Waves 3 and 4 of the Cambridge study when participants were 12–13 and 14–15 years of age, respectively. Externalizing behavior was measured with eight items (disobedient, difficult to discipline, restless, cheats, lies, daredevil, quarrelsome-aggressive, and reacts to criticism or punishment) that were rated by the child’s teacher using a 3-point scale (seldom = 1, sometimes = 2, and frequently = 3) to produce a score that ranged from 8 to 24. Internal consistency was good for both administrations of the externalizing behavior measure (Wave 3, α = .79; Wave 4, α = .86). Reading performance was measured with a single rating of current reading ability provided by the child’s teacher (1 = very much above average; 2 = a little above average; 3 = average for age; 4 = rather below average; and 5 = very poor for age).

Dependent variables

Delinquency, measured at Wave 6, served as the dependent variable in the first set of analyses and consisted of self-reported involvement in the following nine delinquent acts during the past 3 years: joyriding, property damage, receiving stolen property, shoplifting, stealing from slot machines, breaking and entering, stealing from cars, involvement in fights, and carrying a weapon. Each act was rated on a 4-point scale (0 = never; 1 = one or two times; 2 = three or four times; and 3 = five or more times) to yield a total score that ranged from 0 to 27. The delinquency scale displayed adequate internal consistency in the Cambridge study (α = .71).

Adult convictions between the ages of 17 and 24 served as the dependent variable for the second set of analyses. The total number of convictions obtained between Waves 5 and 8 of the Cambridge study (ages 17–24) was tabulated from the central Criminal Record Office in London. The central Criminal Records Office contains a complete record of all relatively serious offenses committed in Great Britain and Ireland. These records extend back to age 10, the minimum age of criminal responsibility in England.

Control Variables

Seven control variables were selected based on their relevance to one or more of the independent, dependent, or mediating variables in this study. These seven variables were collected during Wave 1 of the Cambridge study and included the boy’s age as of November 1964 (123–130 months = 1; 131–137 months = 2; 138–141 months = 3; and 142–158 months = 4), race (British = 1 and non-British = 2), fretful or difficult baby (no = 1 and yes = 2), parental inconsistency (no = 1 and yes = 2), family income (comfortable = 1; adequate = 2; and inadequate = 3), Progressive Mazes IQ (111 or above = 1; 101–110 = 2; 91–100 = 3; and 90 or below = 4), and parental criminal record (none = 1; juvenile conviction = 2; one adult conviction = 3; two adult convictions = 4; three adult convictions = 5; and more than three adult convictions = 6).

Two Wave 1 proxies for the mediator variables were included as control variables in regressions predicting the mediator and dependent variables: Teacher ratings of externalizing behavior at Wave 1 (concentration problems, difficult relationships, and difficult to discipline—mean inter-item r = .26) and sentence reading quotient at Wave 1 (1 = 101–128; 2 = 89–100; 3 = 71–88; and 4 = 70). A prior measure of convictions (convictions at ages 10–11) was included as a control variable in the regression predicting adult convictions. There was no prior measure of self-reported delinquency to serve as a control variable in the regression predicting self-reported adolescent delinquency.

Procedure

Seven of the first eight waves of the Cambridge study were utilized as part of the current investigation. Precursor or pretreatment variables were collected at Wave 1 when the boys were 8–9 years of age and the independent variable (conduct disorder) was collected at Wave 2 when the boys were 10–11 years of age. The first-stage mediators (externalizing behavior and reading performance) were collected at Wave 3 when the boys were 12–13 years of age and the second-stage mediators (externalizing behavior and reading performance) were collected at Wave 4 when the boys were 14–15 years of age. The self-report dependent variable (delinquency) was collected at Wave 6 when the boys were 18–19 years of age and extended back 3 years. The official record’s dependent variable (adult convictions) was assessed between Waves 5 and 8 when participants were 17 to 24 years of age. Because there was no overlap between waves or observation periods, the design of the current study qualifies as prospective. Use of these secondary data for research purposes was approved by the Kutztown University Institutional Review Board.

Statistical Analyses

Data were analyzed with mediated path analysis using cross-lagged mediators that controlled for mediator continuity (i.e., Wave 3 reading performance and Wave 4 externalizing behavior in testing the externalizing-3 → reading-4 relationship, and Wave 3 externalizing behavior and Wave 4 reading performance in testing the reading-3 → externalizing-4 relationship).¹ Three criteria (Walters, 2014c) were used to evaluate the chaining hypothesis: (1) The indirect effect of the putative mediating chain should be significant, (2) the direct effect of the independent variable on the dependent variable should either be nonsignificant or substantially reduced with introduction of the mediators (Muller, Judd, & Yzerbyt, 2005), and (3) the target chain or chains should achieve a stronger effect than one or more comparison chains.

The third criterion was evaluated in this study by contrasting the conduct disorder-2 (CD-2) → externalizing-3 → externalizing-4 → delinquency (target) chain with the CD-2 → reading-3 → reading-4 → delinquency (comparison) chain, the CD-2 → externalizing-3 → delinquency (target) chain with the CD-2 → reading-3 → delinquency (comparison) chain, and the CD-2 → externalizing-4 → delinquency (target) chain with the CD-2 → reading-4 →delinquency (comparison) chain. These comparisons were made using Preacher and Hayes’ (2008) contrast test with bootstrapped confidence intervals. The criterion for a significant indirect effect or significant difference between chains was a bootstrapped confidence interval that did not include zero.

Missing Data

All 411 cases from the Cambridge study were included in the current investigation. Missing data on individual items ranged from 0.0% to 12.2% (parental inconsistency). Over half (59.6%) the sample had no missing data on any of the 17 variables included in this study, 20.0% had missing data on 1 variable, 11.7% had missing data on 2 variables, 6.3% had missing data on 3 variables, and 2.4% had missing data on 4–6 variables. All missing data for this study were handled with Full Information Maximum Likelihood.

Results

Descriptive statistics for and correlations between the 10 control variables (age, race, fretful baby, parental inconsistency, family income, IQ, parental criminal record, Wave 1 externalizing behavior, Wave 1 sentence reading quotient, and Wave 2 convictions), 1 independent variable (conduct disorder at Wave 2), 4 mediator variables (externalizing behavior and reading performance at Waves 3 and 4), and 2 dependent variables (delinquency at Wave 6 and adult convictions at Waves 5–8) are listed in Table 1. Over a third of the correlations were statistically significant at a Bonferroni-corrected α level of .00037 (.05/136 individual correlations). There was no evidence of multicollinearity between the predictor variables used in this study (tolerance = .602–.962; variance inflation factor = 1.040–1.662).

Table 1.

Descriptive Statistics for and Correlations between the Independent, Dependent, Mediating, and Control Variables for this Study.

Variable	n	Wave	M	SD	Range	B	C	D	E	F	G	H	I	J	K	L	M	N	O	P	Q
A. Age	411	W1	2.46	1.1	1–4	0.04	−0.05	−0.06	0.06	−0.06	−0.04	−.20*	−0.05	0	−0.01	0.01	0	−0.11	−0.07	−0.01	−0.08
B. Race	411	W1	1.13	0.34	1–2		−0.04	0.12	0.06	0.04	0.04	0.02	0.05	0.04	0.07	0.02	0.02	−0.06	0.08	0.07	0.09
C. Fretful baby	376	W1	1.22	0.41	1–2			−0.01	−0.03	−0.04	0	−0.03	−0.04	0.03	−0.07	−0.1	−0.04	−0.04	−0.07	−0.06	−0.04
D. Inconsistency	361	W1	1.28	0.45	1–2				0.18	0.05	0.12	0.06	0.05	.29*	0.14	0.05	0.06	0.06	−0.04	0.06	0.09
E. Family income	411	W1	1.92	0.72	1–3					.27*	0.16	.24*	.18*	.19*	0.13	0.1	.24*	.26*	0.14	0.11	.22*
F. IQ	411	W1	2.38	1.12	1–4						0.15	.36*	.22*	0.15	.25*	.19*	.35*	.38*	0.13	0.17	.18*
G. Externalizing W1	402	W1	0.57	0.8	0–3							0.18	0.08	.25*	.33*	.29*	.28*	.22*	0.12	0.13	.23*
H. Sentence reading	377	W1	2.71	1.16	1–4								0.12	0.12	0.16	0.1	.50*	.44*	0.12	0.07	0.09
I. Criminal parent	411	W1	1.68	1.33	1–6									0.06	0.11	0.13	0.17	.22*	.23*	0.09	.35*
J. Conduct disorder	365	W2	1.65	0.91	1–4										.22*	.32*	.19*	.20*	0.08	.27*	.25*
K. Externalizing W3	404	W3	12.31	2.7	8–23											.41*	.22*	.21*	.21*	.25*	.27*
L. Externalizing W4	378	W4	12.7	3.28	8–24												.23*	.35*	.29*	.21*	.36*
M. Reading W3	401	W3	3.31	1.03	1–5													.52*	.15*	0.17	.20*
N. Reading W4	370	W4	3.1	0.97	1–5														0.14	.19*	.27*
O. Delinquency	389	W6	5.14	4.99	0–27															0.09	.42*
P. Child convictions	411	W2	0.04	0.28	0–4																.30*
Q. Adult convictions	395	W5 -7	0.7	1.57	0–10

Note: n = number of nonmissing values; Wave = wave when variable measured; M = mean; SD = standard deviation; range = range of scores in current sample; age = categorical measure of age in months; race = British (1) versus non-British (2); fretful baby = difficult or fretful as a baby (no = 1 and yes = 2); inconsistency = inconsistent parenting (no inconsistency = 1 and inconsistency = 2); family income = family income (comfortable = 1, adequate = 2, and inadequate = 3); IQ = intelligence quotient—progressive matrices IQ (111 or above = 1, 101–110 = 2, 91–100 = 3, and 90 or below = 4); externalizing = classroom externalizing behavior as evaluated by child’s teacher; sentence reading = sentence reading quotient score (1 = 101 to 128, 2 = 89 to 100, 3 = 71 to 88, and 4 = 70); criminal parent = categorical measure of parental criminality; conduct disorder = presence and severity of childhood onset conduct disorder; reading = teacher rating of child’s current reading performance; delinquency = self-reported delinquency over the past three years.

*p < .00037 (Bonferroni-corrected α).

A summary of the results of the 15-variable path analysis of the conduct disorder–adolescent delinquency relationship (excluding control variables) can be found in Figure 1. Although neither cross-lagged correlation was significant, two of the three indirect effects derived from the externalizing mediators were significant, compared to no significant indirect effects for the three chains derived from reading mediators. In addition, the two significant externalizing chains were significantly different from their counterpart reading chains according to the results of the Preacher and Hayes (2008) contrast test (see Table 2). The two significant externalizing chains were subjected to sensitivity testing, the results of which indicated that a covariate confounder would need to correlate .58 with both the mediator and dependent variable to eliminate the mediating effect of external-3 on the first leg of the CD-2 → delinquency relationship, .44 with the mediator and dependent variable to eliminate the mediating effect of external-4 on the second leg of the CD-2 → delinquency relationship, and .51 with the mediator and dependent variable to eliminate the mediating effect of external-4 on the one-mediator CD-2 → delinquency relationship.

Figure 1.

Standardized coefficients for the path model running from childhood conduct disorder to adolescent delinquency. Note: external-3 → reading-4 = .01 (p = .78) and reading-3 → external-4 = .11 (p = .07); paths involving control variables are not shown. *p < .05; **p < .001.

Table 2.

Total, Direct, and Specific Indirect Effects for the Adolescent Delinquency Outcome.

	BCBCI
	Estimate	Lower	Upper
Total Effect	0.301	−0.318	0.901
Direct Effect	−0.085	−0.716	0.532
Specific Indirect Effects (Two Mediators)
CD-2 → external-3 → external-4 → DEL-6	0.044	0.004	0.119
CD-2 → reading-3 → reading-4 → DEL-6	−0.01	−0.056	0.004
Preacher and Hayes Contrast Test	0.054	0.006	0.133
Specific Indirect Effects (W3 Mediator)
CD-2 → external-3 → DEL-6	0.059	−0.005	0.204
CD-2 → reading-3 → DEL-6	0.018	−0.03	0.138
Preacher & Hayes Contrast Test	0.041	−0.063	0.185
Specific Indirect Effects (W4 Mediator)
CD-2 → external-4 → DEL-6	0.304	0.12	0.577
CD-2 → reading-4 → DEL-6	−0.04	−0.17	0.013
Preacher & Hayes Contrast Test	0.343	0.131	0.652

Note: W3 = Wave 3; W4 = Wave 4; CD-2 = conduct disorder at Wave 2; external-3 = externalizing behavior at Wave 3; external-4 = externalizing behavior at Wave 4; reading-3 = reading performance at Wave 3; reading-4 = reading performance at Wave 4; DEL-6 = delinquency at Wave 6; Preacher & Hayes Contrast Test = comparison between the two chains; estimate = point estimate; BCBCI = 95% bias-corrected bootstrapped confidence interval, lower = lower boundary of bias-corrected bootstrapped confidence interval; upper = upper boundary of bias-corrected bootstrapped confidence interval.

Similar results surfaced when these same measures (plus childhood convictions) were used to assess the conduct disorder–adult conviction relationship (see Figure 2). The two-mediator externalizing chain and one-mediator external-4 chain achieved significance as they did in the delinquency analyses, but this time only the latter was significantly different from its reading counterpart (see Table 3). Sensitivity testing revealed a moderate to high degree of robustness in the mediating effects observed between childhood conduct disorder and adult convictions. A covariate confounder, it appears, would need to correlate .58 with both the mediator and dependent variable to eradicate the mediating effect of external-3 on the first leg of the CD-2 → delinquency relationship, .49 with the mediator and dependent variable to eliminate the mediating effect of external-4 on the second leg of the CD-2 → delinquency relationship, and .51 with the mediator and dependent variable to lower the mediating effect of external-4 on the one-mediator CD-2 → delinquency relationship to zero.

Figure 2.

Standardized coefficients for the path model running from childhood conduct disorder to adult convictions. Note: external-3 → reading-4 = .02 (p = .75) and reading-3 → external-4 = .11 (p = .06); paths involving control variables are not shown. *p < .05; **p < .001.

Table 3.

Total, Direct, and Specific Indirect Effects for the Adult Convictions Outcome.

	BCBCI
	Estimate	Lower	Upper
Total Effect	0.232	0.031	0.449
Direct Effect	0.11	−0.086	0.312
Specific Indirect Effects (2 Mediators)
CD-2 → external-3 → external-4 → convict	0.011	0.001	0.036
CD-2 → reading-3 → reading-4 → convict	0.002	−0.001	0.013
Preacher & Hayes Contrast Test	0.008	−0.006	0.034
Specific Indirect Effects (W3 Mediator)
CD-2 → external-3 → convict	0.014	−0.004	0.062
CD-2 → reading-3 → convict	0.003	−0.011	0.041
Preacher & Hayes Contrast Test	0.011	−0.019	0.058
Specific Indirect Effects (W4 Mediator)
CD-2 → external-4 → convict	0.078	0.022	0.169
CD-2 → reading-4 → convict	0.01	−0.004	0.043
Preacher & Hayes Contrast Test	0.068	0.005	0.159

Note: W3 = Wave 3; W4 = Wave 4; CD-2 = conduct disorder at Wave 2; external-3 = externalizing behavior at Wave 3; external-4 = externalizing behavior at Wave 4; reading-3 = reading performance at Wave 3; reading-4 = reading performance at Wave 4; convict = adult convictions between the ages of 17 and 24; Preacher & Hayes Contrast Test = comparison between the two chains; estimate = point estimate; BCBCI = 95% bias-corrected bootstrapped confidence interval, lower = lower boundary of bias-corrected bootstrapped confidence interval; upper = upper boundary of bias-corrected bootstrapped confidence interval.

Discussion

Consistent with the hypotheses tested in this study, the two-mediator externalizing (from conduct disorder to external-3 to external-4 to outcome) and second-stage mediator externalizing (from conduct disorder to external-4 to outcome) chains achieved significance in predicting self-reported delinquency and official convictions using a bias-corrected confidence interval that did not include zero as the criterion for significance. Sensitivity testing revealed that unobserved extraneous variables would need to achieve moderately large to large correlations (Cohen, 1988) with both the mediator and dependent variables to completely eliminate these two mediating effects. The first-stage mediator externalizing chain (from conduct disorder to external-3 to outcome) and all three reading counterpart chains failed to achieve significance in this study. In addition, the two-mediator externalizing and second-stage mediator externalizing chains were significantly different from their reading counterparts in predicting self-reported delinquency and the second-stage mediator externalizing chain was significantly different from its counterpart in predicting adult convictions. Hence, whereas externalizing behavior was found to mediate the relationship between childhood onset conduct disorder and subsequent delinquency and adult crime, academic achievement in the form of teacher-rated reading ability played no causal role in this relationship, despite correlating with childhood onset conduct disorder. Externalizing behavior, on the other hand, correlated with conduct disorder, both dependent measures, and achieved several significant indirect effects. This would seem to suggest that while externalizing behavior may forge a connection between childhood onset conduct disorder and later offending, reading performance is only a marker for conduct disorder with no real causal role in the conduct disorder–later offending relationship.

The results of this study would appear to be congruent with Gottfredson and Hirschi’s (1990) views on low self-control and Krueger, Markon, Patrick, and Iacono’s (2005) views on the externalizing spectrum. This is certainly true with respect to the finding that childhood onset conduct disorder encouraged increased externalizing behavior, as observed and rated by the child’s teacher, and that later childhood externalizing behavior, in turn, encouraged increased mid-to-late adolescent delinquency and early adult criminality. There were other results, however, that were incongruent with both theories. For instance, Gottfredson and Hirschi (1990) contend that relative self-control is set by age 8–10 and Krueger et al. (2005) maintain that disinhibition is primarily, though not exclusively, a function of heredity. In the current study, even after controlling for the child’s level of externalizing behavior at age 8 or 9 years, externalizing behavior at ages 12 to 15 continued to mediate and shape the relationship between childhood onset conduct disorder and subsequent delinquent and criminal behavior. The proximal chaining and self-altering processes observed in this study would appear to be congruent with DeLisi and Vaughn’s (2014) temperament-based theory of antisocial behavior and Walters’ (2014c) recent study showing that childhood temperament impacted future delinquency and crime by way of certain intervening variables, one of which was externalizing behavior.

The results of this study indicate that externalizing behavior requires regular monitoring and assessment. In that externalizing behavior may be less stable than Gottfredson and Hirschi (1990) or Krueger et al. (2005) have hypothesized, it is essential that it be evaluated and reassessed on a regular basis. In concert with their rejection of the psychology of individual differences, Gottfredson and Hirschi (1990) assert that whereas absolute self-control improves with age, relative self-control or a person’s position on the self-control continuum relative to age-mates does not change once the individual reaches the age of 8 or 9 years. Findings from the current study directly contradict this supposition in that externalizing behavior would not have been able to mediate the relationship between age 10–11 conduct disorder and age 15–19 delinquency after controlling for age 8–9 externalizing behavior if externalizing behavior did not change. Because externalizing behavior changes in both an absolute and relative sense, it should be possible to intervene with individuals suffering from externalizing problems or low self-control after age 10. Parent–child interaction therapy designed to improve both parenting skills and the child–parent relationship has been found effective in reducing externalizing behavior in preschool children (Bagner, Sheinkopf, Vohr, & Lester, 2010), school-based behavioral consultation has been found effective in reducing externalizing behavior in primary school children (Wilkinson, 1997), and multisystemic therapy has been found effective in reducing externalizing behavior in severely antisocial adolescents (Ogden & Hagen, 2006). The latter finding suggests that intervention may be successful in reducing externalizing behavior even in seriously delinquent youth as long as vital personal (impulsivity and callousness), environmental (parenting and peer relationships), and developmental (age and maturity level) factors are addressed.

The results of this study also have implications for those who work with conduct disordered and disruptive youth. Whereas low reading achievement may serve as a marker of early conduct disorder and externalizing tendencies there was no evidence in the current study that reading problems mediated the relationship between conduct disorder and later delinquency. Consequently, while low reading achievement can be used to assess youth, interventions designed to improve reading ability should have little impact on future delinquency and adult criminality. Externalizing behavior, by comparison, is not only a marker of current adjustment difficulties but is also capable of predicting future adjustment difficulties. Moreover, it appears to link early conduct disorder to future delinquency and adult criminality just as it appears to link early temperament to subsequent antisocial behavior (DeLisi & Vaughn, 2014; Walters, 2014c). What practitioners can take away from this is an appreciation for how important assessment is in managing serious antisocial youth. In line with Walters’ (2014c, 2015a, 2015b) “worst of both worlds” hypothesis, individuals falling at the upper end of the externalizing spectrum likely have a history of both early antisocial behavior and substance misuse. These issues need to be thoroughly investigated and addressed through intervention, based on the understanding that reserving the most intensive interventions for the highest risk individuals will produce the best results (Lipsey, Wilson, & Cothern, 2000). For high externalizing youth, it is likely that interventions for both serious antisocial behavior and substance misuse will be necessary and that the underlying temperament dimensions for each pattern—fearlessness in the case of antisocial behavior, negative emotionality in the case of substance misuse, and disinhibition in the case of both antisocial behavior and substance misuse—be addressed (Walters, 2014b).

This study has a number of strengths. First, it employed a longitudinal panel design in which data were collected prospectively with no overlap between waves. Second, a number of control variables were included in the analyses. Some of these control variables are among the strongest predictors of delinquency in the Cambridge study—low intelligence, family criminality, poverty, and parental inconsistency (Farrington, 1995)—and several were Wave 1 precursor measures of the mediator variables, and in the case of early adult offending, the Wave 2 precursor of prior convictions. A commonly voiced criticism of longitudinal survey research is that because many, if not all, of the variables are based on participant self-report, shared method variance may explain the observed relationships (Shadish, Cook, & Campbell, 2002). This criticism does not extend to the present study because each set of variables came from a different source. The conduct disorder score was constructed by the interviewer from information provided by one or both parents, the mediators (externalizing behavior and reading performance) were based on ratings from the child’s teacher, and the independent variables came either from self-report (delinquency) or official records (convictions). Hence, the variable relationships observed in this study cannot be attributed to common or shared method variance because, in fact, the independent, dependent, and mediator variables each came from different sources.

The strengths of a prospective study with precursor measures of mediating and dependent variables that can rule out shared method variance as an explanation of current results notwithstanding, there are still potential problems with the external and internal validity of this study. External validity is of concern because the study was conducted on boys from a working class neighborhood in London, England born around 1953 and used data collected in the 1960s and 1970s. One could accordingly question how well the current results generalize to non-English youth, female participants, and more recent cohorts. In addition, several of the variables may suffer from limited internal validity. First, neither the externalizing nor reading precursors were exactly the same as the Wave 3 and 4 externalizing and reading mediators. This is because only a portion of the items from the Waves 3 and 4 externalizing measures were available at Wave 1, and a teacher rating of reading performance was unavailable at Wave 1 so scores on a Sentence Reading measure administered at Wave 1 were used instead. Moreover, there was no Wave 1 or 2 precursor measure of self-reported delinquency like there was for adult convictions. Finally, the teacher rating of reading performance rating provided only a gross estimate of academic achievement and ability and should be considered a major limitation of the current study. Perhaps if a more comprehensive academic assessment had been performed, evaluating both word recognition and reading comprehension skills, different results would have surfaced. It is uncertain how these threats to the internal validity of some of the measures affected the results of this study but it is possible that they may have inflated some of the pathway estimates while deflating others.

It could be argued that all the current study demonstrates is that early externalizing behavior (childhood onset conduct disorder) predicts teacher-rated early adolescent externalizing behavior which, in turn, predicts mid-to-late adolescent (delinquency) and early adult (convictions) externalizing behavior. A similar criticism has been leveled against Gottfredson and Hirschi’s (1990) self-control theory of crime; i.e., that low self-control is nothing more than an alternate term for offending and so the argument that low self-control causes crime is a tautology (Akers & Sellers, 2004). In response to this criticism, I would like to point out that Wave 3 externalizing behavior and reading performance correlated at about the same level with Wave 2 conduct disorder and that Wave 1 externalizing behavior correlated higher with Wave 3 and Wave 4 externalizing behavior than it did with Wave 2 conduct disorder despite being two to four more years further removed from Waves 3 and 4 externalizing behavior than Wave 2 conduct disorder. It should also be noted that item content differed between the three sets of measures, with conduct disorder being a general index of problematic behavior across different settings, externalizing behavior being a more specific measure of teacher-rated noncriminal impulsive and disruptive behavior, and delinquency/convictions being composed exclusively of actual criminal acts. Despite superficial similarities, then, these three measures appeared to be assessing different constructs.

Perhaps more than anything, the current study provides additional support for the chaining model first introduced in Walters’ (2014c) temperament study. The importance of the chaining concept is that it explains how early temperament and early conduct disorder can impact on distal events, like subsequent delinquency, through a chain of more proximal linking variables. Distal events, then, are linked to one another by mediating variables, whether the mediating variables are social capital, as suggested by the cumulative disadvantage construct (Sampson & Laub, 2003), cognition, as suggested by the psychological inertia construct (Walters, 2012), rewards, as suggested by social learning theory (Akers, 2001), or labels, as suggested by symbolic interactionism (Chiricos, Barrick, Bales, & Bontrager, 2007). The biosocial model is another perspective that may help clarify the role of externalizing behavior in the development of complex causal chains. According to the biosocial model, criminal behavior is the result of interacting biological and social factors (Beaver, Wright, DeLisi, & Vaughn, 2008). Externalizing behavior, being a joint function of biological and environmental influences, would seem to lend itself nicely to this perspective. Research, in fact, indicates that effective behavioral interventions for externalizing behavior can have a favorable impact on brain function (Woltering, Granic, Lamm, & Lewis, 2011) and effective psychopharmacological interventions can have a favorable impact on externalizing behavior (Ercan, Basay, Durak, & Ozbaran, 2011). In the name of theory development and evidence-based practice, then, it is imperative that we seek to integrate variables from different domains, regardless of whether these variables are biological, psychological, or sociological in nature.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Note

References

Akers

R. L.

(2001). Social learning theory. In Paternoster

Bachman

(Eds.), Explaining criminals and crime: Essays in contemporary criminological theory (pp. 192–210). Los Angeles, CA: Roxbury.

Akers

R. L.

Sellers

C. S.

(2004). Criminological theories: Introduction, evaluation, and application (4th ed.). Los Angeles, CA: Roxbury Publishing Company.

Bagner

D. M.

Sheinkopf

S. J.

Vohr

B. R.

Lester

B. M.

(2010). Parenting intervention for externalizing behavior problems in children born premature: An initial examination. Journal of Developmental and Behavioral Pediatrics, 31, 209–216.

Beaver

K. M.

Wright

J. P.

DeLisi

Vaughn

M. G.

(2008). Genetic influence on the stability of low self-control: Results from a longitudinal sample of twins. Journal of Criminal Justice, 36, 478–485.

Bennett

K. J.

Brown

K. S.

Boyle

Racine

Offord

(2003). Does low reading achievement at school entry cause conduct problems? Social Science and Medicine, 56, 2443–2448.

Burke

J. D.

Loeber

Mutchka

J. S.

Lahey

B. R.

(2002). A question for DSM-V: Which better predicts persistent conduct disorder—Delinquent acts or conduct symptoms? Criminal Behaviour and Mental Health, 12, 37–52.

Chiricos

Barrick

Bales

Bontrager

(2007). The labeling of convicted felons and its consequences for recidivism. Criminology, 45, 547–581.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.

Cohen

Kasen

Velez

C. N.

Hartmark

Johnson

… Streuning

E. L.

(1993). An epidemiological study of disorders in late childhood and adolescence: I. Age-specific and gender-specific prevalence. Journal of Child Psychology and Psychiatry, 34, 851–867.

10.

Cohn

van Domburgh

Vermeiren

Geluk

Doreleijers

(2012). Externalizing psychopathology and persistence of offending in childhood first-time arrestees. European Child and Adolescent Psychiatry, 21, 243–251.

11.

Costello

E. J.

Mustillo

Erkanli

Keeler

Angold

(2003). Prevalence and development of psychiatric disorders in childhood and adolescence. Archives of General Psychiatry, 60, 837–844.

12.

DeLisi

Neppl

T. K.

Lohman

B. J.

Vaughn

M. G.

Shook

J. J.

(2013). Early starters: Which type of criminal onset matters most for delinquent careers? Journal of Criminal Justice, 41, 12–17.

13.

DeLisi

Vaughn

M. G.

(2014). Foundation for a temperament-based theory of antisocial behavior and criminal justice system involvement. Journal of Criminal Justice, 42, 10–25.

14.

DeLisi

Vaughn

M. G.

Beaver

K. M.

Wexler

Barth

A. E.

Fletcher

J. M.

(2011). Fledgling psychopathy in the classroom: ADHD subtypes, psychopathy, and reading comprehension in a community sample of adolescents. Youth Violence and Juvenile Justice, 9, 43–58.

15.

Ercan

E. S.

Basay

B. K.

Basay

Durak

Ozbaran

(2011). Risperidone in the treatment of conduct disorder in preschool children without intellectual disability. Child and Adolescent Psychiatry and Mental Health, 5, 10.

16.

Farrington

D. P.

(1995). The development of offending and antisocial behaviour from childhood: Key findings from the Cambridge Study in Delinquent Development. Journal of Child Psychology and Psychiatry, 560, 929–964.

17.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

18.

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York, NY: Guilford.

19.

Huizinga

Loeber

Thomberry

T. P.

(1993). Longitudinal study of delinquency, drug-use, sexual-activity, and pregnancy among children and youth in three cities. Public Health Reports, 108, 90–96.

20.

Kenny

D. A.

(2013). Mediation: Sensitivity analysis [webinar]. Retrieved from http://davidakenny.net/webinars/Mediation/Sensitivity/Sensitivity.html.

21.

Krueger

R. F.

Markon

K. E.

Patrick

C. J.

Iacono

W. G.

(2005). Externalizing psychopathology in adulthood: A dimensional-spectrum conceptualization and its implications for DSM-V . Journal of Abnormal Psychology, 114, 537–550.

22.

Lipsey

M. W.

Wilson

D. B.

Cothern

(2000, 4). Effective intervention for serious juvenile offenders. Juvenile Justice Bulletin. Washington, DC: Office of Juvenile Justice and Delinquency Prevention.

23.

Maughan

Pickles

Hagell

Rutter

Yule

(1996). Reading problems and antisocial behavior: Developmental trends in comorbidity. Journal of Child Psychology and Psychiatry, 37, 405–418.

24.

Moffitt

T. E.

(1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674–701.

25.

Mordre

Groholt

Kjelsberg

Sandstad

Margrethe

(2011). The impact of ADHD and conduct disorder in childhood on adult delinquency: A 30-year follow-up study using official crime records. BMC Psychiatry, 11, 57.

26.

Muller

Judd

C. M.

Yzerbyt

V. Y.

(2005). When moderation is mediated and mediation is moderated. Journal of Personality and Social Psychology, 89, 852–863.

27.

Murray

Farrington

D. P.

(2010). Risk factors for conduct disorder and delinquency: Key findings from longitudinal studies. Canadian Journal of Psychiatry, 55, 633–642.

28.

Muthén

(1998–2007). Mplus user’s guide (5th ed.). Los Angeles, CA: Author.

29.

Nock

M. K.

Kazdin

A. E.

Hiripi

Kessler

R. C.

(2006). Prevalence, subtypes, and correlates of DSM-IV conduct disorder in the National Comorbidity Survey Replication. Psychological Medicine, 36, 699–710.

30.

Ogden

Hagen

K. A.

(2006). Multisystemic therapy of serious behaviour problems in youth: Sustainability of effectiveness two years after intake. Child and Adolescent Mental Health, 11, 142–149.

31.

Petersen

I. T.

Bates

J. E.

D’Onofrio

B. M.

Coyne

C. A.

Lansford

J. E.

Dodge

K. A.

… Van Hulle

C. A.

(2013). Language ability predicts the development of behavior problems in children. Journal of Abnormal Psychology, 122, 542–557.

32.

Preacher

K. J.

Hayes

A. F.

(2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, 879–891.

33.

Rolon-Arroyo

Arnold

D. H.

Harvey

E. A.

(2014). The predictive utility of conduct disorder symptoms in preschool children: A 3-year follow-up study. Child Psychiatry and Human Development, 45, 329–337.

34.

Sampson

R. J.

Laub

J. H.

(1993). Crime in the making. Cambridge, MA: Harvard University Press.

35.

Shadish

W. R.

Cook

T. D.

Campbell

D. T.

(2002). Experimental and quasi-experimental designs for generalized causal inference. Boston, MA: Mifflin.

36.

Tracy

P. E.

Wolfgang

M. E.

Figlio

R. M.

(1990). Delinquency in two birth cohorts. Washington, DC: US Office of Juvenile Justice and Delinquency Prevention.

37.

Vaughn

M. G.

DeLisi

Beaver

K. M.

Wexler

Barth

Fletcher

(2011a). Juvenile psychopathic personality traits are associated with poor reading achievement. Psychiatric Quarterly, 82, 177–190.

38.

Vaughn

M. G.

DeLisi

Gunter

Beaver

K. M.

Perron

B. E.

Howard

M. O.

(2011b). The severe 5%: A latent class analysis of the externalizing behavior spectrum in the United States. Journal of Criminal Justice, 39, 75–80.

39.

Vaughn

M. G.

Salas-Wright

C. P.

DeLisi

Maynard

B. R.

(2014). Violence and externalizing behavior among youth in the United States: Is there a severe 5%? Youth Violence and Juvenile Justice, 12, 3–21.

40.

Waldman

I. D.

Lahey

B. B.

(2010). Oppositional defiant disorder, conduct disorder, and juvenile delinquency. In Beauchaine

T. P.

Hinshaw

S. P.

(Eds.), Child and adolescent psychopathology (2nd ed., pp. 411–438). Hoboken, NJ: Wiley.

41.

Walters

G. D.

(2012). Crime in a psychological context: From career criminals to criminal careers. Thousand Oaks, CA: Sage.

42.

Walters

G. D.

(2014a). Crime and substance misuse in adjudicated delinquent youth: The worst of both worlds. Law and Human Behavior, 38, 139–150.

43.

Walters

G. D.

(2014b). Does substance abuse fall exclusively along the externalizing spectrum? Substance Use and Misuse, 49, 724–734.

44.

Walters

G. D.

(2014c). Pathways to early delinquency: Exploring the individual and collective contributions of difficult temperament, low maternal involvement, and externalizing behavior. Journal of Criminal Justice, 42, 321–326.

45.

Walters

G. D.

(2015a). Criminal and substance involvement from adolescence to adulthood: Precursors, mediators, and long-term effects. Justice Quarterly, 32, 729–747.

46.

Walters

G. D.

(2015b). Recidivism and the “worst of both worlds” Hypothesis: Do substance misuse and crime interact or accumulate? Criminal Justice and Behavior, 42, 435–451.

47.

West

D. J.

Farrington

D. P.

(1973). Who becomes delinquent? London, England: Heinemann.

48.

Wilkinson

L. A.

(1997). School-based behavioral consultation: Delivering treatment for children’s externalizing behavior in the classroom. Journal of Educational and Psychological Consultation, 8, 255–276.

49.

Willcutt

E. G.

Pennington

B. F.

(2000). Psychiatric comorbidity in children and adolescents with reading disability. Journal of Child Psychology and Psychiatry, 41, 1039–1048.

50.

Williams

McGee

(1994). Reading attainment and juvenile delinquency. Journal of Child Psychology and Psychiatry, 35, 441–459.

51.

Wolfgang

Figlio

Sellin

(1972). Delinquency in a birth cohort. Chicago, IL: University of Chicago Press.

52.

Woltering

Granic

Lamm

Lewis

M. D.

(2011). Neural changes associated with treatment outcome in children with externalizing problems. Biological Psychiatry, 70, 873–879.

53.

Young

S. E.

Friedman

N. P.

Miyake

Willcutt

E. G.

Corley

R. P.

Haberstick

B. C.

Hewitt

J. K.

(2009). Behavioral disinhibition: Liability for externalizing spectrum disorders and its genetic and environmental relation to response inhibition across adolescence. Journal of Abnormal Psychology, 118, 117–130.