Dyadic Taxonomy of Delinquent Youth

Abstract

Using latent class analysis (LCA), this study identified a dyadic taxonomy of delinquent youth categorized by varying types of maternal-youth reporting discrepancies (i.e., youth < maternal, youth > maternal) within a sample of 764 14-year-old high-risk youth. Four distinctive subgroups of youth were identified, two of which reported more than a minimal degree of informant discrepancy across all domains of delinquent behavior. One subgroup exhibited higher maternal-reported delinquency in comparison with youth reports, and one subgroup exhibited higher youth-reported delinquency in comparison with maternal reports. Additionally, risk factors (e.g., peer delinquency, caregiver monitoring) and delinquency-related difficulties (e.g., police contact) were associated with youth placement in the LCA-identified subgroups. Study findings suggest that youth with higher levels of self-reported delinquency compared with maternal reports may be at greater risk for delinquency-related difficulties due to problems related to inadequate monitoring by caregivers combined with greater involvement with peer delinquency.

Keywords

informant discrepancy taxonomy delinquency

Introduction

Numerous researchers have emphasized the need for data collection from multiple informants to accurately assess delinquent behavior (e.g., Achenbach, 2011; De Los Reyes, 2011; Kraemer et al., 2003; Loeber & Farrington, 2008). However, many studies have found low agreement when comparing responses provided by children, parents, teachers, or peers (De Los Reyes & Kazdin, 2005; Granero, Ezpeleta, Domenech, & Osa, 2008; Smith, 2007). Consequently, questions persist regarding the origins of informant discrepancy, the most effective analytical approach for combining data from multiple informants, and its implications for measurement, theory, and appropriate intervention (De Los Reyes & Kazdin, 2005; Granero et al., 2008; Kraemer et al., 2003). Researchers have considered whether discrepant perspectives between informants during early adolescence may offer useful information related to risk and protective factors or have predictive utility related to youth problems or adjustments (De Los Reyes, 2011; Ferdinand, van der Ende, & Verhulst, 2004, 2006; Goodman, 2013). For instance, maternal-youth reporting discrepancies of delinquent behavior with minimal reporting by mothers may serve as a marker for weakened mother-child nurturing evidenced by mothers’ lack of awareness of their child’s delinquent behavior. This lack of awareness and disclosure may support theories of delinquency based on socialization arguments. Conversely, maternal-youth reporting discrepancies with minimal reporting by youth may indicate youth self-serving cognitive distortions, including denial or minimization of negative behavior that could foster delinquency (Plante et al., 2012; Wilson, Bouffard, & Mackenzie, 2005). This calls to mind perspectives on delinquent behavior that rely on deviant attitudes, beliefs, and neutralizations. Such youth may benefit from targeted intervention such as increased monitoring, feedback about their behavior, and support of positive behavior (Todd, Campbell, Meyer, & Horner, 2008).

In response to these evolving research and therapeutic issues related to informant discrepancies in reporting of youth delinquency, the current study utilized latent class analysis (LCA) to identify a dyadic taxonomy of delinquent youth and explore youth subgroups distinguished by binary patterns of maternal- and youth-reported delinquency within a sample of 764 high-risk 14-year-old adolescents. Also, to gain greater understanding of the various identified subgroups, theoretically relevant covariates and delinquency-related outcomes during early and late adolescence (e.g., police contact, expulsion from school) were also evaluated.

Explanations of Informant Discrepancies

Informant discrepancy generally has been attributed to (a) inherent variation in behavior in different contexts, (b) disparity in attentiveness to specific aspects of behavior based on each informant’s unique role, and (c) individual bias or differential judgment (De Los Reyes, Thomas, Goodman, & Kundey, 2013; Kraemer et al., 2003). For instance, in comparison with parents, teachers may be better reporters of certain externalizing disorders, such as attention deficit hyperactivity disorder, due to their familiarity with age-appropriate behaviors and regular opportunity to observe attention deficit impairment in academic settings (Althoff et al., 2006; Antrop, Roeyers, Oosterlaan, & Van Oost, 2002; Kraemer et al., 2003; Pelham, Fabiano, & Massetti, 2005; Smith, 2007). Caregivers are considered adequate reporters of childhood behaviors, but as children reach preadolescence, self-report data are considered more reliable as parents’ knowledge of their children’s behavior becomes less reliable (De Los Reyes & Kazdin, 2005; Smith, 2007). Aggressive behaviors are expected to be more congruently reported by different informants as they are more easily observable, particularly in comparison with symptoms of internalizing disorders such as depression or anxiety. Conversely, greater reporting discrepancies of aggressive behaviors are expected due to certain informants, either youth or caregivers, being unwilling to report violent and sexual behaviors (Karver, 2006).

As noted, researchers investigating informant discrepancy have begun to more carefully consider whether the existence of discrepant perspectives between informants may offer useful information related to risk and protective factors or provide predictive utility related to youth problems or adjustments (De Los Reyes & Kazdin, 2005; De Los Reyes et al., 2011; Goodman, De Los Reyes, & Bradshaw, 2010). The argument that informant discrepancy, particularly caregiver-youth informant discrepancy, may be instructive aligns with various developmental theories seeking to explain delinquency. For example, age-graded theory of informal social control considers the negative impact of informal family and social control (e.g., weak parental supervision, peer delinquency) on youth behavior (Sampson & Laub, 1993, 2003). Similarly, general strain theory emphasizes the effects of caregiver strain (e.g., maternal depression, lack of social support) on parenting (Agnew, Rebellon, & Thaxton, 2000). As a result of strain, parents may neglect, fail to monitor, and cease to engage with their children, elevating the likelihood of delinquency (Agnew et al., 2000). Drawing from these theories, caregiver-youth informant discrepancy may not only be a product of variation in setting, informant attentiveness, or individual bias—parent-child reporting discrepancies may also be indicative of weakening social connections and heightened risk of delinquency.

Correspondingly, a number of identified predictors of informant discrepancy are also regarded as important risk markers of delinquency such as socioeconomic status (SES), youth intellectual vulnerabilities, and family/caregiver adversity or mental health problems (Farrington, 2005; Lahey, Moffitt, & Caspi, 2003). Race/ethnicity has also often been explored as a predictor of youth delinquency (Peck, 2013; Piquero, Piquero, & Stewart, 2014). Some studies find greater informant discrepancy among raters of African American children compared with Caucasian children (De Los Reyes & Kazdin, 2005). Researchers theorize this may be due to “less consensus among African American mothers and children as to whether a given behavior is problematic” (De Los Reyes & Kazdin, 2005, p. 246). Other risk markers of delinquency such as disadvantaged family environments and caregiver adversity, particularly maternal depression, have been linked to informant discrepancy (De Los Reyes & Kazdin, 2005; Granero et al., 2008; Stokes, Pogge, Wecksell, & Zaccario 2011; Youngstrom, Loeber, & Stouthamer-Loeber, 2000). Relatedly, researchers suggest monitoring may impact informant discrepancy (Ferdinand et al., 2006), yet few studies have examined this link (De Los Reyes, Goodman, Kliewer, & Reid-Quiñones, 2010).

Informant discrepancy has been linked to youth issues such as immaturity or intellectual impairments, resulting in youth not rating their behaviors as problematic (Granero et al., 2008). Relatedly, some researchers suggest that youth depict their problem behavior in a favorable light (i.e., social desirability), while others suggest that youth exaggerate their problems (Stokes et al., 2011). However, the impact of social desirability on informant discrepancy and the direction of the association have rarely been investigated (De Los Reyes & Kazdin, 2005; Stokes et al., 2011). Additionally, no studies have investigated the possibility of an association between caregiver social desirability and informant discrepancy. Yet, parents—especially mothers—report stigma and feeling fearful of being blamed for their child’s delinquent behavior (Aldridge, Shute, Ralphs, & Medina, 2011) that may result in reporting discrepancies. Lastly, while the strong influence of delinquent peers on youth delinquency has been extensively investigated (Pratt et al., 2010), peer delinquency has not yet been explored as a risk factor for reporting discrepancies of delinquency. Associating with delinquent peers during early adolescence has been linked to poor parental monitoring (Dishion, Patterson, Stoolmiller, & Skinner, 1991), with some indications that ties to delinquent peers may decrease disclosure and increase parent-child alienation (Schneider & Younger, 1996), which in turn may increase reporting discrepancies.

Additionally, numerous researchers have asserted that reports of informant discrepancy may be exaggerated by certain analytic strategies utilized to combine informant data (Brown & Barlow, 2005; Kraemer, 2007; Romano et al., 2004). Several researchers have recently noted that LCA can be useful in clarifying the relationship between multiple indicators and an underlying taxonomy of a psychological or behavioral disorder, particularly when using data from different informants (Althoff et al., 2006). Examining parent-youth informant discrepancy of youth victimization, Goodman (2013) identified subgroups of youth by jointly assessing parental- and youth-reported exposure to violence, resulting in a dyadic taxonomy. Similarly, a primary contribution of the current study was the use of LCA to identify a dyadic taxonomy distinguished by binary patterns of maternal- and youth-reported delinquent behaviors.

Current Study

With data from a sample of 764 14-year-old high-risk youth, the current study utilized LCA to identify a dyadic taxonomy of delinquent youth and explore youth subgroups distinguished by binary or dual assessment patterns of maternal- and youth-reported delinquency, clarifying whether youth and mothers over- or underendorse youth delinquent behavior universally or in certain domains (e.g., aggression, property offenses). Based on prior studies, at least two subgroups with informant discrepancy were expected—one with higher youth-reported delinquency compared with maternal reports (youth > maternal) and one with higher maternal-reported delinquency compared with youth reports (maternal > youth) (Ferdinand et al., 2004, 2006). In addition, theoretically relevant risk factors (e.g., caregiver monitoring) and delinquency-related difficulty (e.g., police contact) were examined to provide greater understanding of characteristics of youth subgroups distinguished by patterns of maternal and youth reports and to detect whether reporting discrepancies may be useful in predicting youth adjustment problems.

Method

Sample

This study utilized data drawn from the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) research project (Runyan et al., 2014). This project was comprised of five coordinated longitudinal research projects in the United States using ecological-developmental theory to examine the antecedents and consequences of child maltreatment. The LONGSCAN data were collected between 1991 and 2009. Delinquency items collected from youth and caregivers at the 14-year-old wave available from the National Data Archive on Child Abuse and Neglect (NDACAN) were used for the current study. Demographics, correlates, and outcome measures were collected during the 12-year-old, 14-year-old, and 18-year-old waves. Detailed information regarding the LONGSCAN sample, methods, and measures has been provided elsewhere (Knight, Smith, Martin, Lewis, & The LONGSCAN Investigators, 2010; Runyan et al., 1998). The measures directly related to the current study are fully described here.

Of the full sample of 888 youth participating in the 14-year-old visit of LONGSCAN data collection, 86% (n = 764) included both self-report and maternal report of youth delinquent behavior with inclusion rate varying across the five research sites from 67% to 94%. All youth participating in the study are considered at risk youth or have substantiated exposure to maltreatment. For instance, the sample from the Northwest (n = 156) was comprised of children with substantiated or unsubstantiated cases of maltreatment reported to child protective services (CPS). The sample from the South (n = 157) was developed from a state public health program that identified children at high risk of maltreatment. The sample from the Southwest (n = 168) consisted of children in foster care at age 4, children who were in foster care at one time and eventually adopted by age 4, and children who were in foster care but returned to their families by age 4. The sample from the Eastern region (n = 182) was comprised of children who sought pediatric services at one of three clinics serving low income populations in an inner city setting. The sample from the Midwest (n = 101) was obtained from families reported to CPS who were receiving comprehensive care or treatment. A sample of comparison children was selected from the same neighborhood.

The gender distribution of the sample was 49.0% male and 51.0% female with no significant differences in the gender distribution across study sites, χ²(4) = 3.37, n.s. The ethnic distribution of the sample was comprised of 58.8% African American, 26.4% Caucasian, 5.0% Hispanic, and approximately 9.8% of the subsample reported their ethnicity as biracial or as other than one of the races/ethnicities specified. Ethnic distribution varied by study site, χ²(24) = 300.98, p < .001, with African American youth more prevalent in the East (93.4%), Midwest (68.3%), and South (70.1%) sites while the Southwest and Northwest sites included a lower percentages of African American youth (40.6% and 20.5%). Due to this variation in distribution and prior research indicating higher informant discrepancy between African American children and their caregivers in comparison with Caucasian children, being African American was included as a covariate or predictor in the study models in order to account for site differences and to assess the impact of being African American on informant discrepancy. Mean differences in the total number of maternal- and youth-reported delinquency items were compared across the five study sites. There were no significant differences observed across the sites, F(4, 759) = 1.04, n.s.

Measures of Delinquent Behavior

Youth report and maternal report of delinquent behavior during the 14-year-old visit were collected using the National Institute of Mental Health Diagnostic Interview for Children (NIMH-DISC IV; Shaffer, Fisher, Lucas, Dulcan, & Schwab-Stone, 2000). The NIMH-DISC IV features a computer-assisted interview format designed for use by interviewers without clinical training. Using a computer-assisted program, all interview questions are read by the interviewer verbatim, which greatly minimizes the likelihood of interviewer error. The DISC-Y was administered to the youth while the DISC-P asked the youth’s primary maternal caregiver to provide information regarding her child. These instruments contained 15 delinquent behavior items answered by both youth and caregiver (Table 1). Responses to questions in the DISC-Y and DISC-P used in the current study were limited to yes/no responses.

Table 1.

Descriptive and Bivariate Statistics of Youth Self-Reported and Maternal-Reported Delinquent Behavior (N = 764).

Delinquency items	Youth	Maternal	Endorsement by				Phi coefficient/tetrachoric correlation
Delinquency items	$\bar{X}$ (SD)	$\bar{X}$ (SD)	Neither	Youth only	Maternal only	Both	Phi coefficient/tetrachoric correlation
1. Stolen from family	.15 (.36)	.20 (.40)	71%	9%	14%	6%	.21/.39**
2. Shoplifted	.24 (.43)	.14 (.34)	70%	16%	6%	8%	.30/.52**
3. Stolen from anyone else	.16 (.37)	.11 (.32)	78%	11%	6%	5%	.28/.52**
4. Out >2 hours past curfew	.26 (.44)	.22 (.41)	63%	15%	12%	10%	.25/.42**
5. Ran away	.07 (.25)	.06 (.24)	90%	4%	3%	3%	.48/.80**
6. Lied to get money	.16 (.36)	.17 (.38)	72%	11%	12%	5%	.17/.32**
7. Lied to avoid paying money	.14 (.35)	.15 (.36)	74%	11%	12%	4%	.11/.23**
8. Skipped school	.20 (.40)	.14 (.35)	74%	12%	6%	8%	.40/.65**
9. Broken someone’s things	.10 (.30)	.11 (.31)	81%	8%	9%	2%	.11/.25**
10. Started fires without permission	.11 (.32)	.08 (.28)	83%	9%	6%	3%	.18/.39**
11. Had sexual experience with other	.24 (.43)	.11 (.31)	73%	17%	4%	7%	.34/.61**
12. Bullied someone	.11 (.32)	.19 (.39)	73%	8%	15%	3%	.09/.19*
13. Threatened someone	.10 (.31)	.10 (.30)	82%	8%	8%	3%	.17/.38**
14. Fight with injuries	.39 (.49)	.25 (.43)	52%	23%	9%	16%	.31/.49**
15. Tried to seriously hurt other	.12 (.32)	.04 (.20)	85%	11%	3%	1%	.05/.16
Cronbach’s alpha	.81	.80

p < .05. **p < .01.

Measures of Risk Markers

Based on prior studies, certain youth, caregiver, and familial risk markers were included as covariates in a multinomial logistic regression (MLR). These analyses were undertaken to determine whether they were associated with youth placement in the analytically derived classes distinguished by type of maternal-youth reporting similarities and discrepancies.

Demographics

Demographic measures included Gender (male = 1; female = 2), African American, and family income during the 14-year-old visit.

Social desirability

Youth social desirability was measured during the 12-year-old visit using the LONGSCAN adaptations of items from the Revised Children’s Manifest Anxiety Scale (RCMAS; Knight et al., 2010; C. Reynolds & Richmond, 1997). A subset of items within RCMAS (i.e., “How often do you lie?”/“How often do you tell the truth?”) was originally developed to detect the tendency of respondents to “fake good” in a socially desirable direction and labeled the Lie Scale (Dadds, Perrin, & Yule, 1998). Cronbach’s alpha for LONGSCAN modified six-item youth social desirability or Lie scale was .63.

Similarly, maternal social desirability was measured during the 12-year-old visit using the LONGSCAN adaption of the Marlowe-Crowne (MC) Form A (Knight et al., 2010; W. M. Reynolds, 1982). LONGSCAN researchers modified the answer sets, utilizing a 4-point response set rather than the original 5-point response set, changing the response options from Definitely True to Definitely False, to Strongly Agree to Strongly Disagree. Cronbach’s alpha for the five-item maternal Social Desirability Scale was .68.

Maternal depression

A 20-item scale measuring of depression of primary maternal caregivers was created by LONGSCAN researchers from a pool of items from previously validated depression scales including main components of depressive symptomatology (Knight et al., 2010). Response categories indicate the frequency of occurrence of each item, and are scored on a 4-point scale ranging from 0 (rarely or none of the time) to 3 (most or all of the time). Total scores can range from 0 to 60 with higher scores indicating more depressive symptoms. Cronbach’s alpha for the 20-item maternal depression scale was .99.

Caregiver monitoring

This study included parallel measures designed to assess caregiver’s level of monitoring of youth and youth’s perceived level of monitoring by his or her caregiver (Knight et al., 2010). During the 14-year-old visit, five items were asked of each informant to assess caregiver knowledge of youth’s friends, activities and whereabouts, and money use (e.g., “How much do your parents know about where you are most afternoons after school?”). Variations of these items have been used in multiple studies in which higher levels of parental monitoring have been associated with better adolescent outcomes (Pettit, Bates, & Dodge, 1993). The five items were scored on a 3-point scale ranging from 0 to 2. The items were summed with higher scores indicating higher levels of monitoring. Cronbach’s alpha for the five maternal-reported items was .71 and the Cronbach’s alpha was .72 for the youth-reported items.

Peer delinquency

Youth perception of peer delinquency was measured during the 12-year-old and 14-year-old visit and based on youth response to the 13 items regarding youth’s perception of how many close friends use tobacco, alcohol, and drugs, and engage in risky behaviors. The substance use and risk behavior items were modified from self-report items found in the Youth Risk Behavior and Monitoring the Future surveys (Johnston, O’Malley, Bachman, & Schulenberg, 2010; Knight et al., 2010). Each item was measured on a 3-point scale (0 = none of my friends, 1 = some of my friends, and 2 = most of my friends). The sum of 13 items was used as the measure of youth perception of peer delinquency. Cronbach’s alphas for the 13-item scale during the 12-year-old and the 14-year-old visit were both .83.

Delinquency-Related Correlates

Lastly, associations of delinquency-related difficulties (e.g., police contact, expulsion from school) with the identified subgroups were assessed to validate the analytically identified taxonomy and to determine if certain types of maternal-youth reporting discrepancy of delinquent behavior (i.e., maternal > youth vs. youth > maternal) may be more predictive of later sociolegal difficulties than others. Youth-reported police contact (0 = no, 1 = yes) and maternal-reported police contact (0 = no, 1 = yes) occurring during the past year were reported at the 14-year-old visit based on responses to an item (i.e., been in trouble with police). Similarly, youth-reported school expulsion (0 = no; 1 = yes) and maternal-reported school expulsion (0 = no; 1 = yes) occurring during the past year were reported at the 14-year-old visit. At the 18-year-old visit, information regarding youth lifetime arrest (0 = no; 1 = yes) and below expected level of school attainment for youth (0 = currently in 12th grade, graduated from high school, or obtained General Educational Development [GED]; 1 = currently in <12th grade or dropped out) as reported by mothers was available. These measures were also used to validate the analytically identified taxonomy and to determine if maternal-youth reporting discrepancy of delinquent is predictive of later difficulties.

Analytic Plan

This study utilized latent class analysis (LCA), a person-centered analytic approach that identifies empirically supported taxonomies based on the clustering of responses among manifest indicators (Lanza, Collins, Lemmon, & Schafer, 2007; McCutcheon, 1987; Muthén, 2002). In this study, LCA was used to estimate classes of youth based on binary patterns of youth and maternal responses to 15 indicators of delinquent behavior, exposing dyadic patterns of informant agreement and discrepancy. Other studies investigating informant reporting discrepancies have created separate taxonomies and then compared the likelihood of placement in similar classes by using cross-tabulation of class assignment across taxonomies identified by various informant reports (Althoff et al., 2006; Althoff, Rettew, Ayer, & Hudziak, 2010; Baillargeon et al., 2001; Romano et al., 2004). However, comparison across class placement is vulnerable to high rates of error based on the combination of uncertainties of proper class placement across both models. The inclusion of both mother and youth reports into one dyadic model facilitated higher confidence in proper class placement based on information from two informants. After the identification of the best fitting model, covariates were tested as predictors of youth placement in the subgroups characterized by various types of maternal-youth reporting discrepancies. Usefulness of the taxonomy was assessed through its ability to distinguish youth by police contact and school expulsion.

A series of latent class models was analyzed using Mplus 7 (Muthén & Muthén, 1998-2012). Comparisons of several model fit indicators were made in order to ascertain which latent class model provided the best fit to the observed pattern of responses. Latent class model fit and classification quality were determined by a combination of criteria from several indices, including (a) the Bayesian information criterion (BIC), with lower values indicating a better fit; (b) the “entropy” statistic, which ranges from 0 to 1, with values closer to 1 indicating a clearer delineation of classes; (c) the level of concurrence between the predicted and actual classification of cases in each of the derived classes, with a higher level of agreement indicating better fit; (d) the Lo-Mendell-Rubin (LMR) likelihood ratio test, which compares the estimated model with a model containing one less class; and (e) the bootstrapped likelihood ratio test (BLRT), which is similar to the LMR in assessing whether a given model demonstrates a superior fit compared with a model with one less class by using bootstrapped samples to estimate the log likelihood difference test statistic. Both the LMR and the BLRT provide a p value, indicating whether model fit significantly improves with the inclusion of one additional class (Nylund, Asparouhov, & Muthén, 2007). After the identification of the best fitting model, conditional item probabilities were provided, reflecting the likelihood of youth and maternal endorsement of each delinquency item given youth class placement. Paired sample t tests were used to determine whether there were statistically significant differences between class mean conditional probabilities based on maternal report and youth report for each delinquency item.

In order to gain a fuller understanding of the various subgroups, a preliminary procedure supported by the Mplus program was used to determine which of the possible theoretically relevant covariates were more likely to influence the placement of youth into the LCA-derived classes (Muthén & Muthén, 1998-2012). Next, variables identified as possible predictors of membership in the latent classes were included in an MLR. Key estimates from these analyses are presented as odds ratios. Lastly, the equality of means of police contact and school expulsion were tested across latent classes using pseudo-class-based multiple imputations (Asparouhov, 2007). With the pseudo-class approach, latent class membership is assigned according to the posterior distribution of the sample participants. This procedure is repeated multiple times to account for the uncertainty in latent class membership. Results of the comparison of possible covariates across classes are then combined across the multiple draws using rules derived for multiple imputation of missing data.

Results

Descriptive and Bivariate Statistics

Table 1 includes descriptive and bivariate statistics of the binary delinquency indicators along with the percentages of agreement and disagreement of maternal-youth reporting of each item. According to maternal reports in comparison with youth reports, only two delinquent behaviors were substantially more prevalent among the study sample—stole from family and bullied others. According to youth reports in comparison with maternal reports, some delinquent behaviors were substantially more prevalent—shoplifted, stolen from others besides family, skipped school, involved in fight with injuries, and tried to seriously hurt another. As would be expected, more youth reported having sexual experiences with others than was reported by their mothers. Across the other seven delinquency indicators, the prevalence of youth and maternal reports were the same or very similar (± 3%).

The phi-coefficients (Φ) and tetrachoric correlation coefficients (r tet; Greer, Dunlap, & Beatty, 2003) for the binary maternal-youth endorsement of each delinquency indicator are reported in Table 1. Significant bivariate associations indicate that informant reports of delinquency items were associated at a higher rate than would be expected by chance alone. Positive correlations indicate that if one informant reported the behavior, the other informant was also more likely to report the behavior. The correlations between binary reports were all positive and all were statistically significant except for one item (tried to seriously hurt another). The strength or magnitude of the relationship ranges from 0 to 1, with higher numbers indicating stronger associations. The strongest associations between binary delinquency indicators were found for the items, ran away (Φ = .48) and skipped school (Φ = .40). The weakest associations between the binary delinquency indicators were found for the items, tried to seriously hurt another (Φ = .05) and bullied someone (Φ = .09).

Latent Class Analyses

The LCA process involved specifying a model that included maternal-reported and youth-reported delinquent behaviors (15 delinquency indicators). In deciding on the number of classes that best fit the observed pattern of responses, comparisons were made among several model fit indicators (Table 2). The four-class model appears to provide the best fit to the data. The BIC values decreased with the addition of each class until the five-class specification. The BIC value (BIC = 16,761.34) was lowest for the four-class model, indicating a superior fit in comparison with the other models. The observed value and significance level of the LMR adjusted test for the four-class model (LMR = 272.66, p = .006) suggests that the three-class model can be rejected in favor of the four-class specification. Also, the LMR test for the five-class model (LMR = 151.05, p = .207) indicates that the four-class model cannot be rejected in favor of the five-class specification. Collectively, the available fit indicators tend to converge around the four-class model.

Table 2.

Comparative Model Fit Statistics for Iterative LC Analysis of Youth Based on Youth Self-Reported and Maternal-Reported Delinquent Behaviors (N = 764).

Model	Log likelihood	Bayesian information criterion	Entropy	Lo-Mendell-Rubin adjusted test	BLRT	Mean LC probabilities—Likely class membership
With youth self-report and maternal data
Two-class	−8,387.65	17,180.02	.89	1,941.96 (.000)^a	1,951.40 (.000)^a	.96, .97
Three-class	−8,108.97	16,828.32	.90	554.67 (.000)^a	557.36 (.000)^a	.92, .96, .97
Four-class	−7,972.64	16,761.34	.88	271.34 (.006)^a	272.66 (.000)^a	.96, .90, .90, .95
Five-class	−7,897.11	16,815.95	.86	150.33 (.207)^a	151.05 (.000)^a	.92, .93, .86, .97, .91

Note. LC = Latent class; BLRT = Bootstrapped likelihood ratio test.

H₀: k-1 Class best fit.

The four-class model yielded the following four classes: (a) The smallest class with highest maternal- and highest youth-reported delinquency, labeled High/High class, comprised of 6.3% of the study sample (n = 48); (b) a class with higher maternal-reported delinquency compared with youth-reported delinquency, labeled Maternal > Youth class, that included 20.5% of the study sample (n = 157); (c) a class with higher youth-reported delinquency compared with maternal-reported delinquency, labeled Youth > Maternal class, comprised of 16.1% of the sample (n = 123); and (d) the largest class with the lowest maternal- and lowest youth-reported delinquency, labeled Low/Low class, comprised of 57.1% of the study sample (n = 436).

The first identified class or High/High class had .96 mean probability of proper class assignment, indicating considerable overlap between the expected class assignment based on the model and the actual placement of cases into it. Youth placed in this subgroup possessed moderate to high conditional probabilities across all of the delinquency items indicating extensive involvement in a variety of delinquent behaviors (Figure 1). When comparing conditional probabilities across informants on delinquent behaviors using paired samples t tests, only two pairs of conditional probabilities (i.e., shoplifting and started fires) had statistically significant higher probabilities according to youth report in comparison with maternal report. In sum, this class showed overall agreement between maternal- and youth-reported delinquency with two items displaying higher youth-reported delinquency compared with maternal-reported delinquency.

Figure 1.

Classes with minimal maternal-youth reporting discrepancies of delinquency.

The second identified class or Maternal > Youth class had .90 mean probability of proper class assignment, indicating adequate classification of youth placed in this class. As shown in Figure 2, for this class, conditional probabilities based on maternal report were higher on all of delinquency indicators except the item, had sex with other. When comparing conditional probabilities across informants on delinquent behaviors using paired samples t tests, 10 of 15 items had statistically significant higher probabilities based on maternal report in comparison with those based on youth report. One item, had sex with other, showed statistically significant higher probabilities based on youth report in comparison with maternal report. Statistically significant differences were not found between conditional probabilities for four items (i.e., shoplifted, skipped school, fight with injuries, and ran away). In sum, while this class showed relative maternal-youth agreement on 4 out of the 15 delinquency indicators, maternal-reported delinquency was significantly higher than youth-reported delinquency on 10 items. Additionally, informant discrepancy was not restricted to a certain category of delinquent behavior (i.e., aggression or property offenses) but affected all domains.

Figure 2.

Classes with extensive maternal-youth reporting discrepancies of delinquency.

The third identified class or Youth > Maternal class had .90 mean probability of proper class assignment, indicating adequate classification of youth placed in this class. As shown in Figure 2, for this class, conditional probabilities based on youth report were higher than the conditional probabilities based on maternal report on all of delinquency indicators. When comparing conditional probabilities across informants on delinquent behaviors using paired samples t tests, statistically significant higher estimated probabilities according to youth report in comparison with maternal report were found between all items. Overall, this class showed the greatest informant discrepancy with no items with similar conditional probabilities across maternal and youth reports. Importantly, this class had the highest youth-reported conditional probabilities compared with all other classes on three delinquency indicators related to aggression.

The fourth identified class or Low/Low class had .95 mean probability of proper class assignment, indicating sound classification of youth placed in this class. Youth placed in this subgroup possessed low conditional probabilities across all forms of delinquent behavior indicating relatively low involvement in delinquency with a relatively high degree of maternal-youth agreement (Figure 1). When comparing conditional probabilities on delinquent behaviors using paired samples t tests, 11 out of 15 pairs of conditional probabilities showed no statistically significant difference across informants. On three items (i.e., had sex with other, involved in fight with injuries, and tried to seriously hurt another), statistically significant higher estimated probabilities according to youth report in comparison with maternal report were found. Conversely, for one item (i.e., bullied others), the probability based on mother report was significantly higher in comparison with youth report. In sum, this class showed general informant agreement with four items exhibiting reporting discrepancies.

LCA Correlates

To better understand the LCA-derived classes, MLR was used to examine their association with covariates measuring demographic attributes, family functioning, as well as maternal and youth characteristics to assess whether discrepant perspectives between informants may offer useful information related to risk factors such as weakening familial connections. Due to the high number of previously empirically linked and potentially influential covariates, a preliminary procedure supported by the Mplus program (Muthén & Muthén, 1998-2012) was utilized to determine which of the covariates were more likely to influence the placement of youth in the four classes. Specifically, we used a bootstrapped Wald test for mean differences that is available as an ancillary analysis for LCA (Muthén & Muthén, 1998-2012). From the initial group of covariates (e.g., gender, race, family income, caregiver monitoring, youth and caregiver social desirability, maternal depression, youth perception of peer delinquency), six covariates (i.e., those with statistically significant mean differences across classes) were identified as potentially influential covariates by this initial ancillary testing procedure. The six identified covariates were African American, family income, maternal depression, maternal report of monitoring, youth perception of caregiver monitoring, and youth perception of peer delinquency at age 14.

Next, MLR examined the influence of these covariates on youth placement in the classes using posterior probability-based multiple imputations (i.e., pseudo-class draws).¹ In Table 3, MLR results are presented in the form of odds ratios, which can be interpreted as the odds of being a member of a certain latent class relative to another based on the given attribute (i.e., African American or not) or level (i.e., severity of maternal depression) of each covariate. Of the six covariates included in the model, five covariates were statistically significant, suggesting their possible usefulness in differentiating youth placement in the various classes. Family income did not differentiate youth placement in the various classes. The top section of Table 3 presents the MLR results in odds ratio form with the Low/Low class as the reference group. African American youth had reduced odds of placement in the Maternal > Youth class and the High/High class rather than in the Low/Low class. Higher maternal depression increased the odds of youth placement in the Maternal > Youth class rather than in the Low/Low class. Increased maternal monitoring reduced the odds of placement the Youth > Maternal class and the High/High class rather than the Low/Low class. Increased youth perception of caregiver monitoring reduced the odds of placement in all classes when compared with placement in the Low/Low class. Increased youth perception of peer delinquency elevated the odds of youth placement in the Youth > Maternal class and the High/High class rather than the Low/Low class.

Table 3.

Multinomial Logistic Regression Identifying Predictors of Latent Class Membership.

Predictors	Youth > Maternal class vs. Maternal > Youth class			Maternal > Youth class vs. High/High class			Youth > Maternal class vs. High/High class			High/High class vs. Low/Low class			Maternal > Youth class vs. Low/Low class			Youth > Maternal class vs. Low/Low class
Predictors	Logit	SE	Odds ratio	Logit	SE	Odds ratio	Logit	SE	Odds ratio	Logit	SE	Odds ratio	Logit	SE	Odds ratio	Logit	SE	Odds ratio
African American	.43	.31	1.54	.84	.42	2.32	1.28*	.41	3.60	−1.72**	.40	0.18	−.88**	.24	0.41	−.44	.27	0.64
Family income	−.11	.05	0.90	.14	.06	1.15	0.03	.06	1.03	−0.10	.06	0.90	.04	.04	1.04	−.07	.04	0.93
Maternal depression	−.01	.01	0.99	.01	.02	1.01	0.00	.02	1.00	0.02	.02	1.02	.03*	.01	1.03	.02	.01	1.02
Maternal monitoring	.29*	.10	1.34	.00	.11	1.00	0.29	.11	1.34	−0.34*	.11	0.71	−.34**	.08	0.71	−.05	.10	0.95
Youth perception of caregiver monitoring	−.11	.07	0.90	.14	.09	1.15	0.03	.09	1.03	−0.33**	.09	0.72	−.19*	.07	0.83	−.29**	.07	0.75
Youth perception of peer delinquency	.17**	.04	1.19	−.20**	.05	0.82	−0.04	.05	0.96	0.30**	.05	1.35	.10	.04	1.11	.27**	.04	1.31

Note. Logit = logit coefficients expressing the relationship between covariates and class placement in the logit scale; SE = standard error.

p < .01. **p < .001.

The lower section of Table 3 compares placement in the Youth > Maternal class with placement in the Maternal > Youth class as well as the placement of youth in these two subgroups with the High/High class as the reference group. Two covariates differentiated placement of youth in the two classes with higher informant discrepancy when compared with placement in the High/High class. African American youth were at increased odds for placement in the Youth > Maternal class in comparison with placement in the High/High class. Also, increased youth perception of peer delinquency reduced the odds of placement in the Maternal > Youth class in comparison with placement in the High/High class. Two covariates differentiated placement of youth in the two classes with higher informant discrepancy. Increased maternal monitoring and youth perception of peer delinquency increased the odds of youth placement in the Youth > Maternal class rather than in the Maternal > Youth class.

Lastly, the equality of means of informant-reported police contact and school expulsion at age 14 (youth-reported and maternal-reported) were tested across latent classes using pseudo-class-based multiple imputations (Asparouhov, 2007). The Wald test revealed significant class differences in the youth-reported police contact, χ²(3) = 80.12, p < .001, and maternal-reported police contact, χ²(3) = 38.19, p < .001. Similarly, significant class differences were found in the youth-reported school expulsion, χ²(3) = 18.58, p < .001, and maternal-reported school expulsion, χ²(3) = 22.68, p < .001. As shown in Figure 3, the High/High class had the highest percentage of youth with endorsement by either informant on the delinquency-related correlates and the Low/Low class percentages were the lowest according to both informants. In comparison, endorsement of police contact and school expulsion of two classes with more informant discrepancy were moderate, with greater variability across informants.

Figure 3.

Percentage of youth with delinquency-related difficulties at 14 years of age and 18 years of age by class membership.

Assessing distal outcomes at 18 years of age (maternal-reported), significant class differences were found in youth arrest, χ²(3) = 29.42, p < .001, and in below expected level of school attainment, χ²(3) = 13.23, p < .01. As shown in Figure 3, percentages of youth with arrests and below expected level of school attainment followed a similar pattern to the indicators collected when the youth were 14 years old, with the High/High class containing the highest percentage of youth with arrests and below expected level of school attainment. The Low/Low class had the lowest percentage of youth with arrests and below expected level of school attainment. The two classes with more informant discrepancy fell between the two other classes, with similar percentages of arrests and below expected level of school attainment.

In sum, these findings support the analytically derived taxonomy of delinquent youth. The prevalence of delinquency-related outcomes follows an expected pattern, with a greater percentage of youth within the classes reporting higher levels of delinquency, whether congruently reported or only reported by one informant (mother or youth), also reporting a greater likelihood of encountering delinquency-related difficulties at age 14 and age 18. Comparisons across classes indicate that congruent reporting on delinquent behavior between informants has the best predictive utility of youth difficulties or youth resilience (Baillargeon et al., 2001). Reporting discrepancies on delinquent behavior during early adolescence emerges as a warning sign of youth difficulties persisting into late adolescence.

Discussion

Maternal-youth informant discrepancy of delinquent behavior during adolescence has largely been assumed to be the product of developmental processes with youth disclosing less to caregivers as they age (De Los Reyes & Kazdin, 2005; Smith, 2007). However, few researchers have explored whether informant discrepancy may serve as possible indicator of risk with predictive utility related to youth problems or adjustments (De Los Reyes, 2011; Goodman, 2013). This study sought to build on prior research and add to our current understanding of parent-youth reporting discrepancies by identifying a dyadic taxonomy of delinquent youth containing subgroups of youth categorized by varying types of informant discrepancy (i.e., youth < maternal, youth > maternal) and by clarifying whether youth and/or mothers over- or underendorsed youth delinquent behavior universally or only within certain domains (e.g., aggression, property offenses). Four distinctive subgroups of youth were identified, two of which reported more than a minimal degree of informant discrepancy across all domains of delinquent behavior. As expected, one subgroup exhibited higher maternal-reported delinquency in comparison with youth reports and one subgroup exhibited higher youth-reported delinquency in comparison with maternal reports. In addition, the associations of risk factors and delinquency-related difficulty with the identified subgroups were assessed in order to detect additional distinguishing characteristics of youth placed in each of the LCA-identified subgroups.

Maternal-youth reporting discrepancies were not restricted to one type of delinquent behavior. Rather, statistically significant informant discrepancy was found across behaviors that could be categorized as rule breaking, property offenses, as well as more severe behaviors that cause harm to others. This finding suggests that informant discrepancy is not due to disparity in attentiveness to specific aspects of delinquent behavior by these youth or by their mothers as a similar direction of the discrepancy, either youth underreporting compared with mother or mother underreporting relative to the youth, was consistently observed across various types of delinquent behaviors within the classes. Examination of the influence of theoretically relevant covariates, discussed below, provide several explanations of the underlying processes that may be contributing to the systematic bias in youth and maternal reports.

Characteristics of Subgroups With Low Informant Discrepancy

The placement of 63% of youth in two subgroups with minimal informant discrepancy is similar to findings of other studies using LCA to combine reports by different informants (Althoff et al., 2006; Baillargeon et al., 2001; Romano et al., 2004). Over 50% of youth were placed in a class with minimal to low delinquency similarly reported by both informants. This finding indicates that informant discrepancy is less likely when youth are very minimally involved in delinquent behavior. Nominal differences were found between youth and maternal reports of delinquent behavior, with minor differences emerging in the reports of sexual involvement and aggressive behavior (Figure 1), supporting the notion that youth are least likely to disclose these two categories of behaviors to their caregivers (Karver, 2006). Youth placed in this group were not likely to encounter delinquency-related difficulties with 13% having an arrest by age 18.

Only 6.3% of youth were placed in the class with high delinquency similarly reported by both informants. The relatively low percentage of highly delinquent youth parallels findings of other studies using nationally representative and high-risk samples of male and female youth based on data from one informant such as youth self-report or caregiver report (Hasking, Scheier, & Abdallah, 2011; Jennings & Reingle, 2012; Wiesner & Windle, 2004). Youth placed in this class were less likely to be African American and these youth reported high peer delinquency and poor caregiver monitoring. Mothers of youth placed in this group also reported poor monitoring of their children although they accurately reported information regarding their children’s delinquent behaviors and delinquency-related difficulties. This finding corroborates prior research regarding the negative influence of delinquent peers during early adolescence and its link to poor parental monitoring (Dishion et al., 1991). While these risk factors appear to impact delinquent behavior and youth difficulties (i.e., 48% with arrest by age 18), for this subgroup of youth these risks did not result in extensive reporting discrepancies.

Characteristics of Subgroups With Extensive Informant Discrepancy

Approximately one third (37%) of youth in the study sample were placed in one of the two subgroups with greater degrees of informant discrepancy, the Maternal > Youth Delinquency class or the Youth > Maternal Delinquency class. In a previous review examining studies of maternal-youth reporting discrepancies, nonclinical samples of youth tended to be characterized by a larger proportion of youth > maternal type of reporting discrepancy while clinical samples tended to have a larger proportion of maternal > youth type of reporting discrepancy (Breuk, Clauser, Stams, Slot, & Doreleijers, 2007). The proportions of youth from the study sample who exhibited one of the two divergent types of maternal-youth reporting discrepancy were similar (20.5% vs. 16.1%), allowing for comparison across the two subtypes.

Of the two subtypes of reporting discrepancy, the Youth > Maternal Delinquency subgroup exhibited the greatest degree of informant discrepancy. Importantly, youth in this class exhibited the highest level of youth-reported involvement in three severe behavioral problems related to aggression (i.e., threatened someone, fight with injuries, and tried to seriously hurt someone). Monitoring reported by mothers of youth placed in the Youth > Maternal class was higher or not significantly different from monitoring reported by mothers of youth in all other classes. Paradoxically, youth perception of caregiver monitoring in the Youth > Maternal class was lower or not significantly different from youth perception of monitoring in the other classes. Mothers of youth in the Youth > Maternal subgroup seemed to consider their monitoring sufficient, while their children did not, and this disconnect in monitoring could play a role in their children’s delinquent behavior. Taken together, these findings suggest that underreporting of delinquent behavior by mothers may serve as a marker for weakening mother-child communication evidenced by mothers’ lack of awareness of their child’s delinquent behavior. These findings parallel recent research linking delinquency to mother-child reporting discrepancies of parental monitoring, highlighting that youth whose mothers reported better monitoring in comparison with youth report were more likely to be involved in delinquency (De Los Reyes et al., 2010). In addition, youth placed in the Youth > Maternal class reported similar levels of peer delinquency as those youth placed in the High/High class and higher levels than youth in the Maternal > Youth and the Low/Low subgroup. The characteristics of youth placed in the Maternal > Youth group corroborate conclusions of Laird, Criss, Pettit, Dodge, and Bates (2008), who found that adolescents are more likely to form relationships with delinquent peers and be negatively influenced by the negative peer behavior when they believe their parents do not closely monitor their behavior. In addition, the elevated level of peer delinquency reported by youth in the Youth > Maternal class as well as the High/High class confirms the robust connection between youth perception of peer delinquency and self-reported delinquency (Meldrum & Boman, 2013) regardless of whether mothers reported higher or lower levels of youth delinquency (i.e., with or without informant discrepancy). The elevated prevalence of arrest (33%) and below expected level of school attainment of youth in this class (36%) by age 18 support the notion that informant discrepancy during early adolescence may serve as a marker for later difficulties.

Finally, mothers of the youth in the Maternal > Youth subgroup had higher levels of depression symptoms in comparison with youth placed in the Low/Low class. This finding aligns with prior research regarding the link between maternal depression and informant discrepancy (De Los Reyes & Kazdin, 2005; Granero et al., 2008; Youngstrom et al., 2000) and corresponds to De Los Reyes and Kazdin’s theoretical explanation of this depression-related bias as a product of attribution bias whereby the “parental psychopathology promotes a parent’s perspective for providing information of his or her child’s problems that influences their recall of more negative information of the child’s problems, relative to other informants” (p. 500). Alternatively, while mothers of youth in this class reported better monitoring of their children than mothers of youth placed in the Youth > Maternal class, they reported less monitoring than mothers of youth placed in the Low/Low class. Their children also reported less monitoring than youth in the Low/Low class. Taken together, these findings suggest that while some mothers may have overreported delinquent behavior, they did not respond with adequate monitoring according to both maternal and youth reports. Poor monitoring combined with maternal depression substantiates suppositions of general strain theory that highlight the effects of caregiver strain, in this case depression, on parenting and risk of youth delinquency (Agnew et al., 2000). As with the Youth > Maternal class, the Maternal > Youth subgroup also exhibited an elevated prevalence of arrest (31%) and below expected level of school attainment (41%) by age 18, corroborating the notion that informant discrepancy during early adolescence may serve as a marker for later difficulties.

Null effects of several covariates on youth placement in the various subgroups also add to our general understanding of the issue. First, these findings do not suggest, as other research studies have (De Los Reyes & Kazdin, 2005), that informant discrepancy was higher among African American youth. Importantly, some researchers have also suggested that some youth depict their problem behavior in a favorable light (i.e., social desirability or “faking good”) and this tendency could account for informant discrepancy (De Los Reyes & Kazdin, 2005). However, no link between youth social desirability and youth placement in subgroups with varying degrees of reporting discrepancies was found by this study. Similarly, no link was found between caregiver social desirability and informant discrepancy. In fact, only 5% of mothers in this sample responded to any one of the five items on the social desirability scale in an extreme way, suggesting that these mothers did not tend to give socially desirable responses, thus lending credibility to the validity of their responses as a whole (Beretvas, Meyers, & Leite, 2002).

Limitations

Despite the usefulness of the findings, this study has several limitations. First, the study sample was not nationally representative. The sample was comprised primarily of vulnerable youth who were at risk for maltreatment or who had a history of maltreatment, possibly restricting generalizability of the study findings. Specifically, the composition of the sample could have resulted in a misstatement of the prevalence of youth within the various subgroups or altered the observed item response configurations. However, as previously noted, the distribution of youth—with the largest group of youth having minimal involvement in delinquent behavior and a small group with extensive involvement in delinquency—aligns with other studies using nationally representative and high-risk samples of male and female youth (Hasking et al., 2011; Jennings & Reingle, 2012; Wiesner & Windle, 2004). Second, the primary use of cross-sectional data limited temporal ordering of informant discrepancy and the selected covariates. Therefore, this research was restricted in its ability to provide insight on the cause or progression of informant discrepancy between youth and their mothers.

Selected covariates and youth delinquency-related difficulties at age 18 were provided by the same informants who reported on youth delinquent behavior (i.e., youth and their mothers) and therefore, the study findings may have been influenced by shared informant effects. While it is important to be aware of this limitation, several other informant discrepancies studies also suffer from this problem (Althoff et al., 2010; De Los Reyes et al., 2013; De Los Reyes et al., 2011; Goodman, 2013). As a result of these limitations, it is critical for future studies to incorporate measures from different sources, such as official reports of delinquency, in order to substantiate the study findings. In addition, while this study found that social desirability or youth tendency to depict their problem behavior in a favorable light (i.e., “fake good”) was not linked to informant discrepancy, the social desirability measure used in this study was designed to detect minimization of problem behavior not to detect exaggeration of delinquency. Future research should investigate whether a tendency to exaggerate delinquency (i.e., “fake bad”) to impress peers and establish independence impacts informant discrepancy (Stokes et al., 2011).

Although the analytic technique utilized in this study has advantages relative to a subjectively determined taxonomy of youth based on binary reports, it also has several limitations. First, latent classes have no intrinsic meaning or identity but are assigned labels by researchers and, consequently, are still subject to biases (Kline, 2005). In addition, LCA assumes that between-class differences can adequately account for any variation among individuals in the sample. By making an assumption that there are no within class differences and assigning all members of a class equivalent conditional item probabilities, individual differences may not be fully accounted for in the model (Lanza et al., 2007; McCutcheon, 1987).

Conclusion

Despite these limitations, this study expands the literature regarding informant discrepancy on several fronts. First, this study indicates that the commonly observed caregiver-youth reporting discrepancies of delinquent behavior are not simply a product of measurement error or youth tendency to fake good. By closer examination of informant discrepancy of delinquent behavior, a more comprehensive explanation emerges of the underlying processes generating informant discrepancy. Over 50% of youth were classified as nondelinquent or minimally involved in delinquency according to both youth and their mothers, indicating that informant discrepancy is less likely when youth are very minimally involved in delinquent behavior. While congruent reporting of delinquent behavior by mothers and youth exhibited the best predictive utility of youth difficulties (Baillargeon et al., 2001), reporting discrepancies on delinquent behavior may serve as an indication or warning of potential youth delinquency-related difficulties. As noted by prior researchers, reports of delinquent behavior collected from multiple informants and the extent of informant discrepancy have predictive utility related to youth problems or adjustments (De Los Reyes, 2011; Ferdinand et al., 2004, 2006; Goodman, 2013).

The findings emphasize the importance of attending not only to the extent of informant discrepancy of delinquent behavior but also to the type or direction of the discrepancy (i.e., which informant reports higher levels of delinquent behavior). The study findings suggest that youth who self-reported higher levels of delinquency than their mothers may be at risk for delinquency-related difficulties due to problems related to monitoring by caregivers. Importantly, these youth reported higher involvement in several types of aggressive behavior toward others than all other youth in the sample. These youth perceived reduced monitoring by their caregivers, while their mothers’ reported higher monitoring than other mothers. Taken together, these results suggest that these mothers may be overwhelmed by their child-rearing responsibilities and benefit from targeted interventions related to child monitoring. This type of reporting discrepancy deserves greater attention in future studies. Conversely, maternal depression and reduced monitoring by mothers were linked to maternal overreporting of delinquent behaviors in comparison with youth reports. Due to depression and its debilitating symptoms, caregivers may neglect, fail to monitor, and cease to engage with their children, elevating the likelihood of delinquency (Agnew et al., 2000). Most importantly, these findings emphasize usefulness of gathering information from multiple informants and the need for further investigation of the underlying processes through which informant discrepancy of delinquency is generated.

Footnotes

Authors’ Note

The data used in this publication were made available by the National Data Archive on Child Abuse and Neglect, Cornell University, Ithaca, NY, and have been used with permission. Data from Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) Assessments 0-12 were originally collected by Desmond K. Runyan, Howard Dubowitz, Diana J. English, Jonathan Kotch, Alan Litrownik, Richard Thompson and Terri Lewis & The LONGSCAN Investigator Group. Funding for the project was provided by the Office on Child Abuse and Neglect (OCAN), Children’s Bureau, Administration for Children and Families, Dept. of Health and Human Services. The collector of the original data, the funder, NDACAN, Cornell University and their agents or employees bear no responsibility for the analyses or interpretations presented here.

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.

Notes

Author Biographies

Joan A. Reid is an assistant professor in the Criminology Program at the University of South Florida St. Petersburg. Her research concerns include juvenile victimization, juvenile victim-offender overlap, and mental health treatment. Her recent research has appeared in Crime & Delinquency, Child Maltreatment, and Criminal Behaviour and Mental Health.

Christopher J. Sullivan is an associate professor and director of graduate studies in the School of Criminal Justice at the University of Cincinnati. His research interests include developmental, life-course criminology; juvenile delinquency and prevention policy; and research methodology and analytic methods. His recent work has appeared in Youth Violence and Juvenile Justice, Journal of Youth and Adolescence, and Criminal Justice and Behavior.

References

Achenbach

T. M.

(2011). Commentary: Definitely more than measurement error: But how should we understand and deal with informant discrepancies? Journal of Clinical Child & Adolescent Psychology, 40(1), 80-86. doi:10.1080/15374416.2011.533416

Agnew

Rebellon

C. J.

Thaxton

(2000). A general strain theory approach to families and delinquency. In Fox

G. L.

Benson

M. L.

(Eds.), Families, crime, and criminal justice (pp. 113-138). New York, NY: JAI Press.

Aldridge

Shute

Ralphs

Medina

(2011). Blame the parents? Challenges for parent-focused programmes for families of gang-involved young people. Children & Society, 25, 371-381. doi:10.1111/j.1099-0860.2009.00282.x

Althoff

R. R.

Copeland

W. E.

Stanger

Derks

E. M.

Todd

R. D.

Neuman

R. J.

. . .Hudziak

J. J.

(2006). The latent class structure of ADHD is stable across informants. Twin Research and Human Genetics, 9, 507-522. doi:10.1375/twin.9.4.507

Althoff

R. R.

Rettew

D. C.

Ayer

L. A.

Hudziak

J. J.

(2010). Cross-informant agreement of the dysregulation profile of the Child Behavior Checklist. Psychiatry Research, 178, 550-555. doi:10.1016/j.psychres.2010.05.002

Antrop

Roeyers

Oosterlaan

Van Oost

(2002). Agreement between parent and teacher ratings of disruptive behavior disorders in children with clinically diagnosed ADHD. Journal of Psychopathology and Behavioral Assessment, 24(1), 67-73. doi:10.1023/A:1014057325752

Asparouhov

(2007). Wald test of mean equality for potential latent class predictors in mixture modeling (Technical appendix). Los Angeles, CA: Muthén & Muthén.

Asparouhov

Muthén

(2012). Auxiliary variables in mixture modeling: A 3-step approach using Mplus (Mplus Web Notes: No. 15). Available from www.statmodel.com

Baillargeon

R. H.

Boulerice

Tremblay

R. E.

Zoccolillo

Vitaro

Kohen

D. E.

(2001). Modeling interinformant agreement in the absence of a “gold standard. Journal of Child Psychology and Psychiatry, 42, 463-473. doi:10.1111/1469-7610.00740

10.

Beretvas

S. N.

Meyers

J. L.

Leite

W. L.

(2002). A reliability generalization study of the Marlowe-Crowne Social Desirability Scale. Educational and Psychological Measurement, 62, 570-589. doi:10.1177/0013164402062004003

11.

Breuk

R. E.

Clauser

C. A.

Stams

G. J.

Slot

N. W.

Doreleijers

T. A.

(2007). The validity of questionnaire self-report of psychopathology and parent-child relationship quality in juvenile delinquents with psychiatric disorders. Journal of Adolescence, 30, 761-771. doi:10.1016/j.adolescence.2006.10.003

12.

Brown

T. A.

Barlow

D. H.

(2005). Dimensional versus categorical classification of mental disorders in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders and beyond: Comment on the special section. Journal of Abnormal Psychology, 114, 551-556. doi:10.1037/0021-843X.114.4.551

13.

Dadds

Perrin

Yule

(1998). Social desirability and self-reported anxiety in children: An analysis of the RCMAS Lie Scale. Journal of Abnormal Child Psychology, 26, 311-317. doi:10.1023/A:1022610702439

14.

De Los Reyes

(2011). Introduction to the special section: More than measurement error: Discovering meaning behind informant discrepancies in clinical assessments of children and adolescents. Journal of Clinical Child & Adolescent Psychology, 40(1), 1-9. doi:10.1080/15374416.2011.533405

15.

De Los Reyes

Goodman

Kliewer

Reid-Quiñones

(2010). The longitudinal consistency of mother-child reporting discrepancies of parental monitoring and their ability to predict child delinquent behaviors two years later. Journal of Youth and Adolescence, 39, 1417-1430. doi:10.1007/s10964-009-9496-7

16.

De Los Reyes

Kazdin

A. E.

(2005). Informant discrepancies in the assessment of childhood psychopathology: A critical review, theoretical framework, and recommendations for further study. Psychological Bulletin, 131, 483-509. doi:10.1037/0033-2909.131.4.483

17.

De Los Reyes

Thomas

S. A.

Goodman

K. L.

Kundey

S. M. A.

(2013). Principles underlying the use of multiple informants’ reports. Annual Review of Clinical Psychology, 9, 123-149. doi:10.1146/annurev-clinpsy-050212-185617

18.

De Los Reyes

Youngstrom

E. A.

Pabón

S. C.

Youngstrom

J. K.

Feeny

N. C.

Findling

R. L.

(2011). Internal consistency and associated characteristics of informant discrepancies in clinic referred youths age 11 to 17 years. Journal of Clinical Child & Adolescent Psychology, 40(1), 36-53. doi:10.1080/15374416.2011.533402

19.

Dishion

T. J.

Patterson

G. R.

Stoolmiller

Skinner

M. L.

(1991). Family, school, and behavioral antecedents to early adolescent involvement with antisocial peers. Developmental Psychology, 27, 172-180. doi:10.1037/0012-1649.27.1.172

20.

Farrington

D. P.

(2005). Childhood origins of antisocial behavior. Clinical Psychology & Psychotherapy, 12, 177-190. doi:10.1002/cpp.448

21.

Ferdinand

R. F.

van der Ende

Verhulst

F. C.

(2004). Parent-adolescent disagreement regarding psychopathology in adolescents from the general population as a risk factor for adverse outcome. Journal of Abnormal Psychology, 113, 198-206. doi:10.1037/0021-843X.113.2.198

22.

Ferdinand

R. F.

van der Ende

Verhulst

F. C.

(2006). Prognostic value of parent-adolescent disagreement in a referred sample. European Child & Adolescent Psychiatry, 15, 156-162. doi:10.1007/s00787-005-0518-z

23.

Goodman

K. L.

(2013). Parent-youth discrepancies in ratings of youth victimization: Associations with psychological adjustment. American Journal of Orthopsychiatry, 83(1), 37-46. doi:10.1111/ajop.12010

24.

Goodman

K. L.

De Los Reyes

Bradshaw

C. P.

(2010). Understanding and using informants’ reporting discrepancies of youth victimization: A conceptual model and recommendations for research. Clinical Child and Family Psychology Review, 13, 366-383. doi:10.1007/s10567-010-0076-x

25.

Granero

Ezpeleta

Domenech

Osa

(2008). What single reports from children and parents aggregate to attention deficit-hyperactivity disorder and oppositional defiant disorder diagnoses in epidemiological studies. European Child & Adolescent Psychiatry, 17, 352-364. doi:10.1007/s00787-008-0677-9

26.

Greer

Dunlap

W. P.

Beatty

G. O.

(2003). A Monte Carlo evaluation of the tetrachoric correlation coefficient. Educational and Psychological Measurement, 63, 931-950. doi:10.1177/0013164403251318

27.

Hasking

P. A.

Scheier

L. M.

Abdallah

A. B.

(2011). The three latent classes of adolescent delinquency and the risk factors for membership in each class. Aggressive Behavior, 37(1), 19-35. doi:10.1002/ab.20365

28.

Jennings

W. G.

Reingle

J. M.

(2012). On the number and shape of developmental/life-course violence, aggression, and delinquency trajectories: A state-of-the-art review. Journal of Criminal Justice, 40, 472-489. doi:10.1016/j.jcrimjus.2012.07.001

29.

Johnston

L. D.

O’Malley

P. M.

Bachman

J. G.

Schulenberg

J. E.

(2010). Monitoring the future: National survey results on drug use, 1975-2009. Volume I: Secondary school students (NIH Publication No. 10-758). National Institute on Drug Abuse. Retrieved from http://files.eric.ed.gov/fulltext/ED514370.pdf

30.

Karver

(2006). Determinants of multiple informant agreement on child and adolescent behavior. Journal of Abnormal Child Psychology, 34, 242-253. doi:10.1007/s10802-005-9015-6

31.

Kline

R. B.

(2005). Principles and practice of structural equation modeling (2nd ed.). New York, NY: Guilford Press.

32.

Knight

E. D.

Smith

J. B.

Martin

L. M.

Lewis

, & The LONGSCAN Investigators. (2010). Measures for assessment of functioning and outcomes in longitudinal research on child abuse—Volume 3: Early adolescence (Ages 12-14). Retrieved from http://www.iprc.unc.edu/longscan/

33.

Kraemer

H. C.

(2007). DSM categories and dimensions in clinical and research contexts. International Journal of Methods in Psychiatric Research, 16(S1), S8-S15. doi:10.1002/mpr.211

34.

Kraemer

H. C.

Measelle

J. R.

Ablow

J. C.

Essex

M. J.

Boyce

W. T.

Kupfer

D. J.

(2003). A new approach to integrating data from multiple informants in psychiatric assessment and research: Mixing and matching contexts and perspectives. American Journal of Psychiatry, 160, 1566-1577. doi:10.1176/appi.ajp.160.9.1566

35.

Lahey

B. B.

Moffitt

T. E.

Caspi

(Eds.). (2003). Causes of conduct disorder and juvenile delinquency. New York, NY: Guilford Press.

36.

Laird

Criss

Pettit

Dodge

Bates

(2008). Parents’ monitoring knowledge attenuates the link between antisocial friends and adolescent delinquent behavior. Journal of Abnormal Child Psychology, 36, 299-310. doi:10.1007/s10802-007-9178-4

37.

Lanza

S. T.

Collins

L. M.

Lemmon

D. R.

Schafer

J. L.

(2007). PROC LCA: A SAS procedure for latent class analysis. Structural Equation Modeling, 14, 671-694. doi:10.1080/10705510701575602.

38.

Loeber

Farrington

D. P.

(2008). Advancing knowledge about causes in longitudinal studies: Experimental and quasi-experimental methods. In Liberman

A. K.

(Ed.), The long view of crime: A synthesis of longitudinal research (pp. 257-278). New York, NY: Springer.

39.

McCutcheon

A. L.

(1987). Latent class analysis (Sage University paper series on quantitative applications in the Social Sciences, Series No. 07-064). Newbury Park, CA: SAGE.

40.

Meldrum

R. C.

Boman

J. H.

IV . (2013). Similarities and differences between perceptions of peer delinquency, peer self-reported delinquency, and respondent delinquency: An analysis of friendship dyads. Journal of Criminal Justice, 41, 395-406. doi:10.1016/j.jcrimjus.2013.07.005

41.

Muthén

B. O.

(2002). Beyond SEM: General latent variable modeling. Behaviormetrika, 29, 81-117.

42.

Muthén

L. K.

Muthén

B. O.

(1998-2012). Mplus user’s guide (7th ed.). Los Angeles, CA: Author.

43.

Nylund

K. L.

Asparouhov

Muthén

B. O.

(2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14, 535-569. doi:10.1080/10705510701575396

44.

Peck

J. H.

(2013). Examining race and ethnicity in the context of general strain theory, depression, and delinquency. Deviant Behavior, 34, 706-726. doi:10.1080/01639625.2012.759050

45.

Pelham

J. W. E.

Fabiano

G. A.

Massetti

G. M.

(2005). Evidence-based assessment of Attention Deficit Hyperactivity Disorder in children and adolescents. Journal of Clinical Child & Adolescent Psychology, 34, 449-476. doi:10.1207/s15374424jccp3403_5

46.

Pettit

G. S.

Bates

J. E.

Dodge

K. A.

(1993). Family interaction patterns and children’s conduct problems at home and school: A longitudinal perspective. School Psychology Review, 22, 403-420.

47.

Piquero

N. L.

Piquero

A. R.

Stewart

E. S.

(2014). Sociological viewpoint on the race-crime relationship. In Beaver

K. M.

Barnes

J. C.

Boutwell

B. B.

Noga

Harman

(Eds.), The nurture versus biosocial debate in criminology: On the origins of criminal behavior and criminality (pp. 43-54). Thousand Oaks, CA: SAGE.

48.

Plante

Daigle

M. S.

Gaumont

Charbonneau

Gibbs

Barriga

(2012). Validation of the “How I Think Questionnaire” in a population of French-speaking adolescents with externalizing behaviors. Behavioral Sciences & the Law, 30, 196-210. doi:10.1002/bsl.2001

49.

Pratt

T. C.

Cullen

F. T.

Sellers

C. S.

Winfree

L. T.

Madensen

T. D.

Daigle

L. E.

. . .Gau

J. M.

(2010). The empirical status of social learning theory: A meta-analysis. Justice Quarterly, 27, 765-802. doi:10.1080/07418820903379610

50.

Reynolds

Richmond

(1997). What I think and feel: A revised measure of children’s manifest anxiety. Journal of Abnormal Child Psychology, 25(1), 15-20. doi:10.1023/A:1025751206600

51.

Reynolds

W. M.

(1982). Development of reliable and valid short forms of the Marlowe-Crowne Social Desirability Scale. Journal of Clinical Psychology, 38, 119-125. doi:10.1002/1097-4679(198201)38:1<119::AID-JCLP2270380118>3.0.CO;2-I

52.

Romano

Baillargeon

R. H.

H.-X.

Zoccolillo

Vitaro

Tremblay

R. E.

(2004). A new look at inter-informant agreement on conduct disorder using a latent class approach. Psychiatry Research, 129(1), 75-89. doi:10.1016/j.psychres.2004.06.017

53.

Runyan

D. K.

Curtis

P. A.

Hunter

W. M.

Black

M. M.

Kotch

J. B.

Bangdiwala

. . .Landsverk

(1998). Longscan: A consortium for longitudinal studies of maltreatment and the life course of children. Aggression and Violent Behavior, 3, 275-285. doi:10.1016/S1359-1789(96)00027-4

54.

Runyan

Dubowitz

English

D. J.

Kotch

J. B.

Litrownik

Thompson

, & The LONGSCAN Investigator Group. (2014). Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) Assessments 0-18 [Dataset]. Available from National Data Archive on Child Abuse and Neglect Web site, http://www.ndacan.cornell.edu

55.

Sampson

R. J.

Laub

J. H.

(1993). Crime in the making: Pathways and turning points through life. Cambridge, MA: Harvard University Press.

56.

Sampson

R. J.

Laub

J. H.

(2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press.

57.

Schneider

B. H.

Younger

A. J.

(1996). Adolescent-parent attachment and adolescents’ relations with their peers: A closer look. Youth & Society, 28(1), 95-108. doi:10.1177/0044118x96028001004

58.

Shaffer

Fisher

Lucas

C. P.

Dulcan

M. K.

Schwab-Stone

M. E.

(2000). NIMH diagnostic interview schedule for children version IV: Description, differences from previous versions, and reliability of some common diagnoses. Journal of the American Academy of Child & Adolescent Psychiatry, 39(1), 28-38. doi:10.1097/00004583-200001000-00014

59.

Smith

S. R.

(2007). Making sense of multiple informants in child and adolescent psychopathology: A guide for clinicians. Journal of Psychoeducational Assessment, 25, 139-149. doi:10.1177/0734282906296233

60.

Stokes

Pogge

Wecksell

Zaccario

(2011). Parent-child discrepancies in report of psychopathology: The contributions of response bias and parenting stress. Journal of Personality Assessment, 93, 527-536. doi:10.1080/00223891.2011.594131

61.

Todd

A. W.

Campbell

A. L.

Meyer

G. G.

Horner

R. H.

(2008). The effects of a targeted intervention to reduce problem behaviors: Elementary school implementation of Check In–Check Out. Journal of Positive Behavior Interventions, 10(1), 46-55. doi:10.1177/1098300707311369

62.

Vermunt

J. K.

(2010). Latent class modeling with covariates: Two improved three-step approaches. Political Analysis, 18, 450-469. doi:10.1093/pan/mpq025

63.

Wiesner

Windle

(2004). Assessing covariates of adolescent delinquency trajectories: A latent growth mixture modeling approach. Journal of Youth and Adolescence, 33, 431-442. doi:10.1023/B:JOYO.0000037635.06937.13

64.

Wilson

D. B.

Bouffard

L. A.

Mackenzie

D. L.

(2005). A quantitative review of structured, group-oriented, cognitive-behavioral programs for offenders. Criminal Justice and Behavior, 32, 172-204. doi:10.1177/0093854804272889

65.

Youngstrom

Loeber

Stouthamer-Loeber

(2000). Patterns and correlates of agreement between parent, teacher, and male adolescent ratings of externalizing and internalizing problems. Journal of Consulting and Clinical Psychology, 68, 1038-1050. doi:10.1037/0022-006X.68.6.1038