The relation between behavioral problems and social competence: A correlational Meta-analysis

While some children easily navigate social encounters, other children lack the ability or motivation to interact with peers and adults in a positive way. Social impairment can be displayed in different ways. Disruptive children seem to struggle with emotion regulation, internalization of rules, are slow to develop empathy and conscience, and often lack adaptive problem solving skills [74]. For example, children may interact with others in an aggressive and disruptive way, and over time, social impairment increases the likelihood of rejection and disliking by peers [22]. Thus, social impairment seems to have short and long-term ripple effects encompassing poor social- and behavioral adjustment. Social competence deficits thus often limit the possibilities for future interactions and limit further skill development [35].

Typically, the ability to negotiate age-appropriate social encounters is measured by either social competence or social skills. These constructs tend to be overlapping and somewhat ambiguously defined, as both often cover indices that relate to language, intelligence, attitude, and interaction with the environment [5]. In addition, both constructs are characteristically dynamic as children’s social competence and social skills develop over time. That is, as children age, they encounter new developmental tasks with increasing complexity, and their social interplay gradually moves from social interaction with parents and siblings at home, to spending more time with peers in kindergarten and at school, where peer relations are considered most important in meeting unique and complex social skills [11]. However, to date, there is no agreed upon definition of social competence and social skills, which has resulted in an ongoing controversy about the nature, conceptualization and measurement of these constructs (e.g., Nangle, Grover, Holleb, Cassano, & Fales, 2000; Rose-Krasnor, 1996).

Social competence is a broadly adaptive characteristic and can be defined as “the ability to take another’s perspective concerning a situation and to learn from past experiences and apply that learning to the ever-changing social landscape” (page 1, [67]). In essence, the definition captures a child’s awareness of how one’s behavior affects his/her surroundings and sensitivity to the needs of others. As such, manifestations of social competence are multifaceted and may include friendship, popularity with peers, positive self-concept, social assertion, and so on ([23, 71]. Although the distinction between social competence and social skills is sometimes overlooked, social skills generally refer to specific abilities or behaviors that are needed to perform a task [52]. Social skills describe the ability to accurately select relevant and useful information from a social context, and use this information to explore opportunities for goal attainment and maintain good relationships with others. Social skills are both cognitive and interpersonal, verbal and non-verbal conditions for appropriate social behavior and positive social interactions [5].

Altogether, social competence could be thought of as a more general and evaluative term, whereas social skills are more situation specific behaviors [32] or responses [60]. However, although individual skills contribute to overall social competence, no single behavior is sufficient for social competence (Hupp, LeBlanc, Jewell, & Warnes, 2009), and social competence in one situation not necessarily transfers to other situations [52].

Pervasive and persistent behavioral problems in children and youth are considered as a risk factor for successful functioning in different arenas; at home, at school, and among peers [29, 47]. Furthermore, these problems are also associated with detrimental future consequences, including antisocial behavior, social exclusion, and severe psychopathology ([16, 57]. In a diagnostic approach, conduct disorder and oppositional defiant disorder are the most prevalent diagnoses (DSM-V; ICD-10). However, since there are no corresponding diagnoses for social competence deficits, comparisons between conduct disorders and social competence becomes impossible. We have therefore excluded studies that exclusively use a diagnostic approach.

Research suggests that behavioral problems as early as age three may predict affiliation to a population segment representing high future costs in adulthood [16]. Studies on social competence - behavioral problems relation may also vary according to the range of the problem construct being measured. In this study, we limit our focus on three different types of behavioral problems: externalizing behaviors, conduct problems, and aggression. The terms externalizing behaviors (i.e., negative behaviors that are directed toward the external environment [12];) and antisocial problems are often used synonymously, yet the externalizing behaviors is often used to describe less severe disruptive and destructive child behavior [26, 68]. Conduct problems, on the other hand, are generically defined as high rates of aggression, noncompliance, oppositional behaviors [76], and typically assessed with the Achenbach’s Child Behavior Checklist [1, 46]. Finally, although there are no definite agreement about the definition of aggression [37], it is often considered a subset of broader concepts (e.g., externalizing behaviors and conduct problems [26];). Typically, aggression covers physical and verbal behavior directed towards individuals with an intent to harm (e.g., pushing, kicking, and threatening; APA, 1994). Most studies of aggression have focused on physical aggression, although other types have also been described, like for example indirect [15] and relational aggression [21] that reflect different forms and functions of aggression ([14, 50]. Studies show that childhood aggression is a strong predictor of adult crime and violence [26, 56]. In the current article, the concepts of externalizing behaviors, conduct problems and aggression represent independent measures that cover from a broad to a narrow range of behavioral problems that may possibly correlate differently with social competence.

To summarize, social competence in childhood has become an area of interest for both researchers and clinicians, because of its negative correlation with behavioral problems (or vice versa), and the important role it seems to play in shaping future adjustment abilities in youth and adulthood. From a clinical perspective, effective treatment and prevention of behavioral problems in children should be comprehensive and address both risk and protective factors. That is, on the one hand, risk factors for the development of these problems should be reduced, and on the other, efforts at promoting social competence should simultaneously be encouraged.

The relation between social competence and different manifestations of behavioral problems is, however, not obvious, given the variation in the strength of this relation among numerous studies (e.g., Burt, Obradović, Long, & Masten, 2008 [30, 64];). Due to lack of research examining moderating variables that may explain variations among studies, a meta-analysis is warranted. The main aim of this study was to review and systematically examine studies testing the correlation between behavioral problems and social competence among children, as well as to investigate the role of various moderators of the reported correlations. With this aim, we made a distinction between social competence and social skills, and between types of behavioral problems as externalizing behaviors, conduct problems, and aggression. Furthermore, we investigated if the variations in correlations would be moderated by characteristics of study design and participants. We limited our analysis to samples of preschool and school age children, when problem behavior is more likely to be malleable, and can be successfully changed through interventions programs. Therefore, the relation between behavioral problems and social competence is especially interesting for this age span to enhance our knowledge and develop interventions.

We conducted a search in Medline, PsycINFO and ERIC (all via OVID) in June 2018. The search strategy included the terms social competence, social skills, and interpersonal. These were combined by the boolean operator ‘AND’ with terms describing antisocial, problem behavior, aggression or conduct disorder, as well as terms describing children or adolescents (see search strategies in Additional file 1).

The search was limited to papers published in 2008 or later, and restricted to papers in English, in Scandinavian languages or in Turkish. We conducted a manual search of the references in the studies we identified. Finally, in order to access unpublished data, we contacted researchers in the field by e-mail for data “in press” or for unpublished data. Applying the inclusion criteria below, we examined titles, abstracts or full text for relevance.

For inclusion in the present meta-analysis, studies had to meet the following criteria: (1) reporting correlation and sample size for the relation between social competence and conduct problems, (2) for children and adolescents 3 > age < 13, (3) measured social competence or social skills, (4) measured antisocial behaviors, externalizing behaviors, conduct problems, or aggression, (5) included reports by parent(s) and/or teacher(s), (6) conducted between January 2008, and January 2018, (7) published in English, Scandinavian languages or Turkish. Accordingly, studies were excluded from the analysis if correlations (1) were based on several instruments combined, (2) included pooled responses from different respondents (e.g., average correlation of self and parent reports) (3) were from studies focused on children with co-occurring problems such as ADHD, learning disabilities, and anxiety.

The search resulted in 518 potentially studies, whereof 54 were identified as independent and relevant studies, in accordance with the inclusion criteria. See Table 1 for a list of the included studies with selected descriptives. Each of the selected studies was coded at least twice with the following variables: (1) publication year, (2) correlation and sample size (N), (3) type of social competence measure: whether the study reported social competence or social skills, (4) type of conduct problems: whether the study reported on externalizing behaviors, conduct problems, or aggression, (5) type of instrument: whether the same instrument used to measure both constructs or not, (6) type of respondent: parent vs. teacher or parent and teacher as a dyad, (7) mean age: the average age of participants, in addition as a categorical moderator, i.e., age < 6 and age ≥ 6 (school-age), (8) gender as percentage of females, (9) country of study origin: USA vs. Europe vs Asia and (10) socio-economic status of the sample. When fewer than five studies reported data on a moderator, the variable was excluded from further analysis. Consequently, we were not able to perform country of study analysis on studies from Australia (N = 2), Canada (N = 1), or Africa (N = 1). We were able to retrieve data from three unpublished studies ([44]a, b [73];), and thus analysis was not performed on published vs non-published data. In order to prevent violation of independence of observations (i.e., including data from the same sample several times), the correlation of the first assessment was included in the analysis, when correlations for several time-points were reported. However, the second wave of observation was used in cases when the first assessment was based on children of age two or below, in line with the inclusion criteria. Further, in order to represent all categories, we made some choices, i.e., parent-teacher reports were prioritized before other reports, aggression reports were used over other reports, and conduct problems reports were used before externalizing behaviors reports.

To ensure reliable coding of the moderators, the first two authors (SH and SK) together generated a coding system for the moderators as well as other study characteristics (e.g., type of the outcome measure) and coded each study separately. Interrater agreement was calculated by dividing the total number of congruent observations to the total number of observations and multiplied times 100. Interrater agreement rate was 98.5%. Finally, the two first authors held consensus meetings, to resolve inconsistencies by consulting the article or by discussion, to reach 100% agreement.

Fifty-four non-overlapping samples were identified, and included in the present meta-analysis. The study selection process is presented in Fig. 1.

The characteristics of all studies were displayed in Table 1.The total sample size was 46,822 participants (range: 53–15,070), with a mean number of 867 participants per study. The majority of samples came from USA (N = 27, 50%), followed by Europe (N = 16, 29%), Asia (N = 7, 13%), Australia (N = 2, 4%), Canada (N = 1, 2%) and Africa (N = 1, 2%). Mean age of children was 5.62 years (SD = 1.97), and the percentage of girls ranged from 23.7 to 63.2% (M = 47.37, SD = 9.62).

Figure 2 shows the forest plot with effect sizes and confidence intervals for each sample (k = 54), in addition to the pooled result, calculated with the random effects weights. The average corrected correlation between social competence and behavioral problems was negative and significant (r = − .42 [95% CI = − .48,-. 37], Z = − 12.79, p < .01). That is, overall higher levels of social competence were associated with lower levels of behavioral problems. However, the variation in correlations across samples was significant and high (Q = 2529.01, p < .001) and the I² statistics showed that 97.90% of the observed variability was beyond what could be expected by chance. This not only justifies using the random effects model, but more importantly, it indicates that it is likely to be one or more variables that moderate the relation between social competence and behavioral problems.

The funnel plot for correlations was close to symmetrical, but indicated that 5 studies were missing to the left of the mean. In a trim and fill analysis these five studies were added, which adjusted the overall correlation to r = −.45, [95% CI = − .50,-. 40]. The adjusted correlation was still significantly different from zero, suggesting that publication bias is not affecting the results considerably. Following [36], removing the samples with the highest (Nix et al., 2016; r = −.78) and the lowest (Wilson, Havighurst, & Harley, 2012; r = −.03) effect sizes did not reduce the Q statistic with more than 50% (Q = 2416.68), nor did removing the study with the largest sample size (Ansari, 2018; N = 15,070, Q = 1776.22). Consequently, all of the 54 study samples were included in the meta-analysis.

Due to the heterogeneity of the overall effect sizes in the relation between social competence and behavioral problems, moderator analyses were conducted. Method-related moderators and demographic moderators were analyzed. Table 2 presents the examined categorical moderators and test statistics.

Our findings suggested that whether studies used the broadly defined social competence construct, or the more narrowly defined social skills, did not change the magnitude of the effect size. That is, assessment of social competence vs. social skills was not able to explain variations in the strength of the correlation among studies. However, this finding may reflect the present controversy pinpointing the often-poor discriminatory power in the assessment between these two constructs. That is, constructs are often measured with overlapping or similar items that capture such as communication and interpersonal skills. Although context and indices together combine to form the different constructs of social competence and social skills, these are, by definition, thought to give diverse outcomes. Social skills in one setting do not necessarily imply social skills in another setting, or general social competence [45, 52]. However, several theories of social competence depict a hierarchical relation between social competence and social skills. Gresham and Elliott [33] suggest that social skills and adaptive behavior represent two subdomains under the construct of social competence. Although social competence is considered as a more overarching term, it is also measured in a context (e.g., by teachers at school), and may as such be difficult to separate from social skills. The present results indicate that it does not matter whether one relies on social competence or social skills, at least when studying how these constructs relate to conduct problems. Nevertheless, this result does not dismiss the fact that researchers should be precise and conscious about what construct they are using. It may be that whether one relies on social competence or social skills make a difference in relation to other child behaviors.

The specificity or range of behavioral problems, also, did not moderate the correlation between social competence and these problems. That is, whether externalizing behavior, conduct problems, or aggression were measured, it did not have a significant influence on the magnitude of the effect size. One possible explanation may relate to the high correlation between the various constructs measuring these problems [10], especially in older children. Even though many young children, especially boys, may express normative aggression ([3, 19, 38]; Nærde, Ogden, Janson, & Zachrisson, 2014), a small number of children express persistent and high levels of aggressive behavior later on [63]. These children also show an increased risk for behavioral problems, as aggression takes a broader form. However, most studies in this analysis reported the measure of externalizing problems, while fewer reported on aggression and conduct problems. Thus, it may be that the results would be different, if we had more reports for the two latter constructs. Overall, present results indicated a robust moderate negative correlation between social competence and behavioral problems, independent of the bandwidth of constructs.

Furthermore, some measurement instruments assess both social competence and behavioral problems using the same format across constructs (e.g., HCSBS [53]; or SSIS [34];). In the present study, however, most studies reported using different instruments when assessing social competence and behavioral problems (N = 35). The results indicated that type of instrument did not moderate the overall negative correlation, thus the instrument format does not have an influence on the results.

The type of respondents appeared as a significant moderator of the social competence - behavioral problems correlation. Our results suggested that this correlation was lower when the two constructs were reported by different respondents, compared to same reporter for the both constructs. That is, the relation was low and non-significant when parents reported on behavioral problems and teacher reported on social competence, whereas high and significant when parents or teachers reported on both constructs (r = −.47 for parent reports, and r = −.51 for teacher reports, both significant at p < .001). The low correlation between parent-teacher as a dyad (r = −.18, p < .001) probably reflects the fact that parents and teachers evaluate behaviors in different contexts, at home and at school, respectively. This may result in inconsistencies of perceptions of the behaviors of the same child. Moreover, parents tend to take an ipsative approach when making their evaluations, whereas teachers do normative assessments, i.e., teachers evaluate a child compared to other children. Additional analyses revealed marginal and a non-significant difference between reports of parents and teachers, when both constructs were reported by the same respondent. Hence, the results suggest highly reliable reports of the same respondent, in line with previous studies ([2, 43, 48, 65]. However, still we cannot know whether the same teacher or the same parent reported on both constructs, i.e., whether assessments are highly context dependent rather than person dependent. Furthermore, considering the possible common method variance, our results should still be interpreted with caution.

Percentage of girls in each study was investigated as a continuous moderator on the social competence-behavioral problems relation. Meta-regression analysis did not reveal significant differences for percentage of girls between studies. However, several findings suggest that girls tend to display lower levels of behavioral problems and somewhat higher levels of social competence, compared to boys [10, 61]. Hence, results reflect no gender difference in the strength of the correlation, since both genders will display different levels of social competence and behavioral problems (i.e., girls with high levels of social competence and low levels of behavioral problems versus boys with low levels of social competence and high levels of behavioral problems).

We also examined age as a dichotomous variable, separated as under or over school age (age < 6 vs. age ≥ 6). Age appeared as a significant moderator of the social competence - behavioral problems relation, with a significantly higher correlation among the oldest children. This finding may have several explanations. First, it may reflect that social competence is more easily evaluated and observed among older children, as developmental factors are critical to the assessment of social competence. Furthermore, it may also reflect that measurement instruments are more suitable to capture social competence in older children. Few instruments are developed for children in preschool years, as it is not typically until preschool years that children start engaging in sustained play with other children. Moreover, lack of social competence and behavioral problems are more visible, and have larger consequences, as the child reach school age, which allows children to navigate through different activities and interaction with peer with a fair amount of autonomy [6]. In addition, during school day extracurricular activities are more structured than in preschool years, involving larger peer groups and more coordinated play, in which poor social competences and behavioral problems, become a larger and more visible challenge.

Finally, we investigated whether the mean effect size varied based on study origin. A total of twenty-seven studies were from USA, whereas sixteen studies were European and seven studies were Asian (studies from Canada, Africa and Australia were excluded). Results showed that study origin could not explain variation in the magnitude of the correlation, which imply that the relation between social competence and behavioral problems can be generalized across these continents.

The current meta-analysis summarized the various studies on the relation between social competence and behavioral problems among children and adolescents, and tried to answer why the results are heterogeneous in the literature. By using the bandwidth of constructs, we tried to demonstrate whether the strength of this relation varies across constructs or measures used. The results of this meta-analysis provide additional support for targeting multiple areas, rather than just focusing on behavioral problems in targeted interventions.

Several limitations of this study should be considered. First, although we made an effort to identify all eligible studies using different search methods, it is still possible that some studies were not identified by our search strategy. Second, given our defined child age span, we excluded studies that used self-report measures. Self-report studies would also be more relevant when studying older children. The study did not include studies where social competence was rated by parent and behavioral problems by teachers, since these studies comprised less than five studies, which made it not feasible to study their effect as a moderator. Third, the correlations between social competence and behavioral problems may have been influenced by a third variable, whereof internalizing problems may represent a strong candidate [28]. Moreover, the methodological quality of the studies was not considered beyond the fact that all were rated by the authors to have a satifactory quality for inclusion in the study.