Prediction of Children’s Academic Competence From Their Effortful Control, Relationships, and Classroom Participation

Abstract

The authors examined the relations among children’s effortful control, school relationships, classroom participation, and academic competence with a sample of 7- to 12-year-old children (N = 264). Parents and children reported on children’s effortful control, and teachers and children reported on children’s school relationships and classroom participation. Children’s grade point averages (GPAs) and absences were obtained from school-issued report cards. Significant positive correlations existed between effortful control, school relationships, classroom participation, and academic competence. Consistent with expectations, the teacher–child relationship, social competence, and classroom participation partially mediated the relation between effortful control and change in GPA from the beginning to the end of the school year. The teacher–child relationship and classroom participation also partially mediated the relation between effortful control and change in school absences across the year.

Children’s academic competence is central to their future success. The importance of successfully navigating the challenges of the school environment is highlighted by findings that academic competence is a significant correlate of positive mental health and high school graduation (Caspi, Elder, & Bem, 1987; Ensminger & Slusarcick, 1992). Despite the importance of school success, 15% of adults report that they have not completed high school (Stoops, 2004). Although the majority of research on school success has focused on curricula, structure, teacher–child ratios, and intelligence, there is an increased awareness of the important roles children’s regulatory abilities, school-related relationships, and classroom participation play in contributing to their academic competence. Indeed, Blair (2002) noted that some longitudinal evidence indicates that social and emotional factors relate to aspects of school success or failure even when controlling for general intelligence at school entry.

The literature considering indices of children’s regulatory abilities and measures of academic competence is growing, as is the literature on relational and motivational correlates of school success. However, studies that bridge these literatures are rare. The current study begins to fill this gap and was designed (a) to test if effortful control (an index of regulatory abilities) predicts changes in academic competence (i.e., grades and absences) across a school year; (b) to test if students’ relationships with teachers and peers, as well as their classroom participation, predict changes in academic competence; and (c) to test if relationships and classroom participation partially mediate the relation between effortful control and academic competence. Simultaneously considering constructs from traditionally different areas of research may clarify if and how children’s regulatory abilities predict their academic competence.

We used effortful control (EC) as an index of children’s regulatory abilities. EC is defined as “the efficiency of executive attention—including the ability to inhibit a dominant response and/or to activate a subdominant response, to plan, and to detect errors” (Rothbart & Bates, 2006, p. 129). Children high in EC are believed to be able to voluntarily control their attention and behavior as needed. EC is measured in a variety of ways, often with measures of attentional regulation, persistence, and the ability to delay gratification, as well as with indices of the ability to voluntarily inhibit or activate behavior (Kochanska, Murray, & Harlan, 2000; Rothbart & Bates, 2006). EC processes are linked to children’s emotion-related regulation, and they modulate emotional reactivity and behaviors (Rothbart & Bates, 2006).

Relations Between Children’s Effortful Control and Academic Competence

Several investigators have argued that emotional competence and processes involving executive attention are important for academic success (Blair, 2002; Raver, 2002). Huffman, Mehlinger, and Kerivan (2000) hypothesized that children’s regulatory abilities contribute to competence beyond measures of IQ. In one study, 60% of teachers reported that being sensitive and not disruptive represented important aspects of academic readiness (Lewit & Baker, 1995). Children high in EC likely have many of these skills, do not easily divert from tasks (Zimmerman, 1998), and process detailed situations more accurately than do their peers low in EC (Lemerise & Arsenio, 2000; NICHD Early Child Care Research Network, 2003).

Some evidence supports the hypothesis that components of EC (e.g., attentional regulation, persistence, or delay of gratification) are positively related to reading, math, and linguistic abilities as well as teachers’ reports of competence (Fabes, Martin, Hanish, Anders, & Madden-Derdich, 2003; NICHD Early Child Care Research Network, 2003), although EC sometimes does not relate to grade retention (Willson & Hughes, 2006). Findings that preschoolers’ delay of gratification predicts future verbal intellectual ability and SAT scores provide some evidence that the relations between regulatory abilities and academic competence persist over time (Rodriguez, Mischel, & Shoda, 1989; Shoda, Mischel, & Peake, 1990). Moreover, some data indicate that the relations between EC and academic competence are similar cross-ethnically. For example, Valiente, Lemery-Chalfant, and Castro (2007) found that Mexican American children’s EC was related to teacher-reported academic competence and absenteeism.

The preceding review indicates that children’s EC is positively related to measures of academic competence; however, not all findings are consistent or significant effects are reduced in magnitude when control variables (e.g., sex or socioeconomic status; SES) or correlates of academic competence are added to the models. In addition, the relations generally account for a modest amount of the variance, suggesting that other constructs are also important when considering academic competence. On the basis of the broader literature, it seems likely that students’ relationships and classroom participation may mediate the relations between EC and academic competence.

Relations Between Children’s Relationships and Academic Competence

As noted above, in addition to EC, the relationships children develop and maintain in school have been associated with their academic competence. A supportive teacher–child relationship may buffer children from some risk factors associated with poor performance, perhaps because teachers are more likely to provide extra assistance to children with whom they have a positive relationship (Resnick et al., 1997). Consistent with this hypothesis, Hamre and Pianta (2001) predicted that a high-quality teacher–child relationship motivates teachers to invest extra resources that can promote children’s school success. In contrast, a conflictual teacher–child relationship may increase stress for the child that may interfere with learning and motivation. Longitudinal data suggest that declines in the nurturant teacher–child relationship precede declines in achievement, and there is evidence that teacher-reported negativity in the teacher–child relationship is related to achievement test scores even when controlling for verbal IQ (Hamre & Pianta, 2001; Midgley, Feldlaufer, & Eccles, 1989).

Positive relationships with peers also contribute to children’s academic achievement (Raver, 2002). Indeed, components of social competence such as peer acceptance and friendships are hypothesized to promote social inclusion in the classroom, which may yield resources that foster interpersonal and academic success (Ladd, 2003). Welsh, Parke, Widaman, and O’Neil (2001) found that positive social skills were associated with academic compe tence, and evidence supports the premise that peer acceptance and general levels of social competence are linked to performance in math, reading, and language (see Ladd, 2003, for a review).

The preceding review suggests that both children’s EC and relationships are related to academic competence. In addition, some data support the hypothesis that EC is related to children’s relationships, and this is necessary for relationships to mediate the relation between EC and academic competence. When children are low in EC and disruptive, they are at increased risk of developing poor relationships with their teachers and receiving low levels of instruction (Berndt & Keefe, 1995; Birch & Ladd, 1997; Murray & Greenberg, 2000). This may contribute to negative perceptions of the classroom, which interfere with motivation for subsequent learning and performance (Wentzel, 1999). Consistent with this line of reasoning, children who lack social skills are viewed as difficult to teach and receive low levels of positive feedback (Arnold, Homrok, Ortiz, & Stowe, 1999; McEvoy & Welker, 2000; Shores & Wehby, 1999), but studies that directly measure EC and examine the hypothesized relations are needed.

There is support for the hypothesis that EC is positively related to indices of social competence. For example, children high in EC are often rated high in compliance (Kochanska, Coy, & Murray, 2001; Kochanska, Murray, & Coy, 1997), sympathy (Eisenberg & Fabes, 1998; Valiente et al., 2004), and social competence (Eisenberg, Gershoff, et al., 2001; see Rothbart & Bates, 2006, for a review). In contrast, children low in EC likely have difficulty modulating negative emotions, complying with others’ requests, and avoiding conflictual peer interactions. In summary, there is evidence that academic competence is associated with both EC and children’s relationships, but it is not clear if EC provides unique prediction of academic competence or if relationships partially mediate the EC and academic competence associations.

Relations Between Children’s Classroom Participation and Academic Competence

Some findings indicate that students’ classroom participation is associated with their grades and absences and that classroom participation might partially mediate the relations between EC and academic competence. Theorists argue that participation may reflect an internal motivation and learning-goal orientation that directs one’s behavior toward classroom tasks and demands (Dweck, 1989; Finn, 1993; Gottfried, Fleming, & Gottfried, 1994). In a national report on educational statistics, Finn (1993) noted that students who rarely participate in their classrooms are at risk to perform poorly beyond risks associated with race, ethnicity, language, or family income. Furthermore, scholars suggest that motivation contributes to academic outcomes because it directs students’ actions and activities, perhaps because they are motivated to pursue goals valued in the school context (Wentzel, 1999; Wig-field, Eccles, Schiefele, Roeser, & Davis-Kean, 2006). Evidence indicates that measures of engagement such as classroom participation and school liking are related to academic progress, math skills, language skills, and attendance (Ladd, Birch, & Buhs, 1999; Valiente et al., 2007). Children low in engagement are likely to have difficulty following rules and capitalizing on learning opportunities that are often correlated with indices of cognitive functioning (Bronson, Tivnan, & Seppanen, 1995; Hughes & Kwok, 2006).

After reviewing the motivation-to-succeed literature, Eccles, Wigfield, and Schiefele (1998) noted that “the highest priority in this area is attention to the influence of emotions on motivation” (p. 1075). Although progress has been made (Wigfield et al., 2006), Eisenberg (2006) suggested in the introduction to the Handbook of Child Psychology that much more work is necessary. Relations between EC and classroom participation might be expected, because children must modulate emotions and demands that occur in school environments to remain engaged (Alexander & Entwisle, 1988). Consistent with this line of reasoning, Valiente et al. (2007) found that EC was positively related to school liking (an index of engagement), and school liking mediated the relation between EC and academic competence.

The Present Study

Findings from several often disparate literatures support the premise that children’s EC, relationships with teachers and peers, and classroom participation represent important correlates of academic competence. However, it is unclear if these variables offer unique or overlapping prediction of academic competence. Our first aim in this study was to test if EC predicted measures of academic competence (i.e., grades and absences) at the conclusion of a school year (i.e., spring) while controlling for academic competence at the beginning of the school year (i.e., fall). Our second goal was to test if prediction from EC remained significant after adding measures of children’s relationships and a measure of classroom participation to the model. We expected positive zero-order relations among children’s EC, the teacher–child relationship, social competence, classroom participation, and GPA; the reverse pattern of relations was expected for school absences. In an effort to extend the literature and on the basis of some empirical evidence (Valiente et al., 2007) and theory (Eisenberg, Sadovsky, & Spinrad, 2005), we predicted that the teacher–child relationship, social competence, and classroom participation would mediate the relations between EC and academic competence.

The extant literature indicates that measures of children’s SES often relate to children’s performance in school and attendance. For example, income was positively related to both the classroom environment (NICHD Early Child Care Research Network, 2005) and students’ achievement (Davis-Kean, 2005). In a review, Haveman and Wolfe (1995) concluded that poverty limits students’ academic competence and academic years completed. Nevertheless, as noted above, low classroom participation places a child at risk for school failure regardless of family income (Finn, 1993). To examine the influence of SES, we added SES to all models and expected SES to be positively related to grades and negatively related to absences.

The literature also suggests that children’s sex is associated with their academic competence; however, findings regarding sex are somewhat less consistent than findings for SES. Whereas sex differences in math and reading sometimes fail to reach significance (Davis-Kean, 2005; Simpkins, Davis-Kean, & Eccles, 2006), some research suggested that boys outperformed girls on math tasks (Frome & Eccles, 1998; Jordan, Kaplan, Oláh, & Locuniak, 2006), and other reports found that girls achieved better reading performance than boys (Frome & Eccles, 1998). Although sex differences such as these are generally small, because of these relations, we added children’s sex to all equations when testing our hypotheses.

In addition, because 47% of the sample in this study was Mexican American and the next largest ethnic group represented was European American (30%), we tested if the strength of the relations differed for Mexican American and European American children. Findings from previous work indicate that Latino students perform more poorly in reading and math than do their European American peers (Children’s Action Alliance, 1999; U.S. Department of Education, 2004).¹ Latino students also tend to be absent, to be tardy, and to drop out of school prior to graduation more often than students of other ethnicities (Finn, 1993; U.S. Bureau of the Census, 2004).

Despite some mean level differences, and although much of the data in this area stems from European American children, it seems likely that EC, school relationships, and classroom participation are also important correlates of Mexican American children’s academic competence. There is some evidence that prediction of developmental outcomes from measures of EC in other cultures (e.g., Indonesia and China) is similar to prediction of outcomes found in U.S. samples (Eisenberg, Pidada, & Liew, 2001; Zhou, Eisenberg, Wang, & Reiser, 2004). Furthermore, the relations between U.S. minority children’s regulatory abilities and their socioemotional and school functioning are similar to relations found in studies that include mostly European American children (Blair, 2002; Gilliom, Shaw, Beck, Schonberg, & Lukon, 2002; Schultz, Izard, & Ackerman, 2000). Thus, we did not expect the relations among the study variables to be different for Mexican American versus European American children.

Method

Procedure and Participants

Participants were recruited from two schools in a southwestern U.S. city. Before the study began, parents (N = 561) received an introductory letter informing them of enrollment and participation procedures. In addition, research assistants were available in the schools during parent–teacher conferences to enroll parents and answer questions about the study. To increase the reliability of the construct scores and to reduce shared method variance, we used a multireporter (child, primary caregiver parent, and teacher) method to assess the key constructs. The questionnaires sent to parents assessed children’s EC and social competence. Parent packets were available in either English or Spanish, because some parents (n = 40) preferred to complete their packets in Spanish. A translation and back-translation method was used, and the original English version was compared with the back-translated version to determine equivalence. The process of translation and back translation continued until all differences were resolved. Teachers (N= 22; teachers reported on an average of 12 children) reported on the teacher–child relationship, children’s social competence, and classroom participation. Children reported on their EC, the teacher–child relationship, and their classroom participation. During the school day, a research assistant read all items to children in their classroom. All questionnaires were completed between March and April, and participants were compensated for their participation.

Parents of 122 boys and 142 girls (47% of those eligible) provided consent for themselves and for their children to participate. Children were between the ages of 7 and 12 years (M = 9.57 years, SD = 1.04), attended 1 of 22 regular education classrooms, and provided assent. Forty-seven percent of the parent and child participants were Mexican American, 30% were European American, 5% were African American, 8% were Native American, and 10% were of other ethnic origins. To reduce the heterogeneity of the sample, we did not include children who were solely enrolled in special education services and did not spend any time in regular education classrooms in this study.

The sample represents the sex and ethnic population composition of the classrooms (i.e., the population of the classrooms was 48% boys and 52% Mexican American, 34% European American, 8% African American, and 6% Native American children), demonstrating that our sample of children closely resembles all of those eligible. To further compare those children whose parents provided consent with those whose parents did not, we examined the absence rate, a percentage rate published by the Arizona Department of Education, of those who participated with those eligible. The absence rate in our sample (4%) closely represents the absence rate for those eligible (5%). Because public access of GPA data is unavailable, we were not able to conduct a similar comparison for GPA.

Children were predominantly from two-parent homes (80%) in which the primary caregiver was the child’s biological mother (84%). Seventeen percent of primary caregivers (percentages for secondary caregivers [75% were fathers] are in parentheses; 17%) had less than a high school education, 26% (37%) had a high school education, 28% (19%) had some college education, 16% (11%) had a 2-year-college or trade degree, 7% (7%) had a 4-year-college degree, and 6% (9%) had attended graduate school. Family income ranged from below $15,000 to above $150,000 per year and had a mean range of $30,000 to $50,000 per year.

Measures

SES

Income, primary caregivers’ education, and secondary caregivers’ education were highly related (rs > .45, p < .01); therefore, after standardizing, the items were averaged to form a measure of SES (α = .76).

Effortful control

Children and parents reported on children’s EC. Items from the Attention Shifting (e.g., “I am [Your child is] good at keeping track of several different things that are happening around me [him/her]”), Activation Control (e.g., “If I have [Your child has] a hard assignment to do, I get [he/she gets] started right away”), and Inhibitory Control (e.g., “When someone tells me [your child] to stop doing something, it is easy for me [him/her] to stop”) subscales from the Early Adolescent Temperament Questionnaire (Capaldi & Rothbart, 1992) assessed children’s EC. Items were rated on a 5-point Likert-type scale (1 = almost always untrue to 5 = almost always true), and the average of the items (18 for parents and 16 for children) served as a composite of parent-reported and child-reported EC (αs = .89 and .73 for parents and children, respectively). Capaldi and Rothbart (1992) reported that scores from the scale have good discriminant validity, and there is evidence that parent-reported EC and observed indices of EC are significantly related (Eisenberg, Gershoff, et al., 2001; Valiente et al., 2003).

Teacher–child relationship

Teachers reliably (α = .90) rated 20 items on a 5-point scale (1 = definitely not true to 5 = definitely true) from the Student-Teacher Relationship Scale to report on the closeness and conflict of the teacher–child relationship (e.g., Hamre & Pianta, 2001; Pianta, 2001). This scale has been used often in studies with elementary school children, and scores have demonstrated internal consistency. Children rated their relationship with their teacher on a 3-point scale (α = .92; 1 = not at all to 3 = a lot of the time) using an age-appropriate version of the Student–Teacher Relationship Scale. Evidence that the measure correlates in the expected directions with later academic performance, attitudes, and involvement supports the convergent validity of the Student–Teacher Relationship Scale (Birch & Ladd, 1997; Hamre & Pianta, 2001).

Social competence

Two subscales of a slightly modified (Eisenberg, Fabes, Guthrie, & Reiser, 2000) version of Harter’s (1982) Perceived Competence Scale for Children were used to assess children’s social competence. Socially appropriate behavior was the average of four items (e.g., “This child is usually well behaved”; αs = .87 for parents and .85 for teachers). Popularity was the average of three items (e.g., “This child has a lot of friends”; αs = .88 for parents and .89 for teachers). Parents’ reports of socially appropriate behavior and popularity were highly correlated, r(217) = .73, p < .001, as were teachers’ reports, r(236) = .53, p < .001. Thus, the scales were averaged within reporter to form separate measures of parents’ and teachers’ reports of social competence.

Classroom participation

Teachers used 11 items from the Teacher Rating Scale of School Adjustment (Birch & Ladd, 1997; Ladd, Kochenderfer, & Coleman, 1996) to rate children’s classroom participation. Items were rated on a 3-point scale (0 = doesn’t apply to 2 = certainly applies). Teachers’ reports of children’s classroom participation (e.g., “This child follows instructions,” “This child challenges him/herself to do well in school”) were reliable (α = .94). Children reliably (α= .67) rated items on an age-appropriate version (e.g., “I follow my teacher’s instructions”) of this measure.

Academic competence

Official school records were used to obtain measures of children’s academic competence. At the conclusion of the school year, we obtained records of full school days missed and tardies from the fall and spring quarters of the school year. Because the number of full school days missed and tardies were significantly related at the first and last quarters (rs = .24 and .27, ps < .001, respectively), we standardized the number of full school days missed and tardies and then averaged the standardized scores. In the remainder of the article, we refer to this composite as absences. Consistent with Pierce, Hamm, and Vandell (1999), we averaged scores in language, vocabulary, and math (all rs > .60) to form children’s fall and spring GPAs (1 = a grade of F to 5 = a grade of A).

Results

Prior to hypothesis testing, we computed a series of preliminary analyses to test for potential age and sex differences. Next, we examined zero-order relations among the study variables. Finally, mixed model regressions were computed to test the hypotheses. We concluded by testing if the strength of the findings was different for Mexican American versus European American participants.

Complete data were available for 77% of the participants. To avoid problems associated with listwise deletion (see Schafer & Graham, 2002), we imputed missing values using the expectation maximization algorithm after specifying a normal distribution with the missing value analysis program in SPSS Version 12.0. Little’s missing completely at random test was not significant, χ²(288) = 291.60, ns, which supports this method of imputing missing data. Because the patterns of findings were similar for imputed and nonimputed data, we present the results obtained on the single imputed set.

Preliminary Analyses

Table 1 contains the means and standard deviations for the study variables. To examine sex differences, we computed separate multivariate analyses of variance (MANOVAs) by reporter. There were significant multivariate effects (Hotelling’s T) for child-reported measures, F(3, 260) = 5.86, p < .001, and for teacher-reported measures, F(2, 261) = 18.17, p < .001. Univariate tests indicated that teachers reported closer teacher–child relationships with girls than with boys and rated girls higher than boys in classroom participation, Fs(1, 262) = 25.02 and 34.22, ps < .001, respectively. Girls reported higher levels of EC, closer teacher–child relationships, and more classroom participation than boys reported, Fs(1, 262) = 13.56, 9.36, and 5.05, ps < .001, .01, and .05, respectively. In addition, parents rated girls higher in EC than they did boys, t(263) = 21.60, p < .001. There was also a significant multivariate effect for grades and absences, F(4, 259) = 7.26, p < .001. Univariate tests indicated that girls performed significantly better academically than boys did in fall and in spring, Fs(1, 262) = 21.82 and 24.94, ps < .001, respectively; however, there were no significant sex differences in fall or spring absence patterns.

Table 1.

Descriptive Statistics for Study Variables

Variable	M	SD	Range
Socioeconomic status	3.28	1.37	0–7
Effortful control: Child report	3.41	0.52	2.00–5.00
Effortful control: Parent report	3.27	0.60	1.11–4.72
Social competence: Teacher     report	3.05	0.68	1.00–4.00
Social competence: Parent     report	3.00	0.78	1.00–4.06
Teacher–child relationship:     Child report	2.21	0.44	1.00–3.00
Teacher–child relationship:     Teacher report	3.90	0.61	1.50–5.00
Classroom participation: Child     report	2.34	0.31	1.20–3.00
Classroom participation:     Teacher report	2.44	0.51	1.18–3.00
GPA (fall)	3.92	0.87	1.00–5.00
GPA (spring)	3.89	0.94	1.00–5.00
Absences (fall)	0.97	1.12	0.00–6.50
Absences (spring)	1.86	1.78	0.00–12.00

Note. Statistics for socioeconomic status and absences are presented prior to standardizing scores. The standardized scores are used in all other analyses.

Table 2 presents the correlations among all the variables and illustrates the similar pattern of findings for imputed versus non-imputed data. As shown in Table 2, when considering the imputed data, 63 of 78 correlations were significant. Irrespective of reporter, the measures of EC, the teacher–child relationship, social competence, and classroom participation were all significantly related to children’s GPAs in fall and spring. EC, both child and parent reported, was positively related to children’s social competence (parent and teacher reported), teacher–child relationship (child and teacher reported), and classroom participation (child and teacher reported). The indices of children’s EC, teacher–child relationship, and social competence were significantly related to children’s absences from school. The zero-order correlations provide initial support for the hypotheses.²

Table 2.

Zero-Order Correlations for Study Variables

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13
1. Socioeconomic status	—	.22**	.14*	.14*	.35**	−.02	.13	.11	.17*	.27**	.28**	−.07	.05
2. Effortful control: Child     report	.26**	—	.39**	.33*	.26**	.32**	.34**	.32**	.42**	.36**	.41**	−.13*	−.17*
3. Effortful control: Parent     report	.16*	.41**	—	.35**	.26**	.27**	.20**	.16*	.43**	.46**	.49**	−.08	−.09
4. Social competence:     Teacher report	.18*	.35**	.37**	—	.28**	.19**	.65**	.15*	.77**	.38**	.46**	−.16*	−.24**
5. Social competence: Parent     report	.36**	.27**	.27**	.29**	—	.05	.24**	.16*	.30**	.32**	.32**	.06	.13* -
6. Teacher–child relationship:     Child report	−.03	.33**	.29**	.20**	.05	—	.30**	.51**	.27**	.22**	.30**	−.11	.17*
7. Teacher–child relationship:     Teacher report	.14*	.35**	.19**	.65**	.25**	.31**	—	.26**	.70**	.29**	.33**	−.13*	−.21**
8. Classroom participation:     Child report	.12*	.34**	.16*	.16*	.17**	.52**	.28**	—	.30**	.22**	.25**	−.07	−.15*
9. Classroom participation:     Teacher report	.20*	.43**	.45**	.78**	.30**	.29**	.70**	.32**	—	.56**	.61**	−.16*	−.28*
10. GPA (fall)	.29*	.38**	.48**	.43**	.32**	.27**	.33**	.26**	.59**	—	.65**	−.12*	−.08
11. GPA (spring)	.27*	.42**	.52**	.48**	.31**	.30**	.35**	.27**	.63**	.67**	—	−.15*	−.23**
12. Absences (fall)	−.10	−.14*	−.08	−.15*	.09	−.10	−.13*	−.08	−.16*	−.12*	−.14*	—	.42**
13. Absences (spring)	.04	−.17**	−.11†	−.24**	.15*	−.17**	−.21**	−.16**	.28**	−.13*	−.23**	.42**	—

Note. Correlations with imputed data are below the main diagonal. Correlations based on pairwise deletion are above the main diagonal.

^†p < .10.

^*p < .05.

^**p < .01.

Table 2 also contains the within-construct relations. Consistent with previous research, child and parent reports of children’s EC correlated .41, and teacher and parent reports of social competence correlated .29. In addition, teacher and child reports of the teacher–child relationship were correlated .31, and their reports of classroom participation were correlated .32. Therefore, because reports of the same construct across reporters were always significant (ps < .01) and because significant relations were found across reporters, to reduce the number of analyses, we created composite scores by averaging across reporters.³ The remainder of the analyses use these composites.

Regression Analyses

The observations in the present study are clustered (i.e., children are nested within classrooms); thus, prior to testing our hypotheses, we examined the intraclass correlations associated with each model. Clustered data can result in a correlation among responses from the same classroom, and standard errors and consequent significance tests may be biased if the correlation is ignored. Hox (2002) considered intraclass correlations values of .05, .10, and .15 to be small, medium, and large, respectively, but also demonstrated that even intraclass correlations of .10 can bias results. The intraclass correlations in this study ranged from .02 to .14, and the average intraclass correlation was .08 (see Table 3 and Table 4 for the intraclass correlation associated with each mixed model regression). Although the intraclass correlations were generally small, because clustering can potentially bias significance tests and resulting conclusions, the remainder of the analyses were computed using mixed models in SPSS Version 12.0, with classroom as a random effect.

Table 3.

Prediction of GPA and Absences From EC

Predictor	β	SE β	R2	ICC
Predicting spring GPA from EC				.07
Fall GPA	.54***	.06
Sex	.14†	.08
SES	.11*	.06	.47
EC	.51**	.10	.51
Predicting spring absences from EC				.02
Fall absences	.43***	.06
Sex	.01	.09
SES	.12*	.06	.19
EC	−.23*	.09	.21

Note. EC = children’s effortful control; SES = family socioeconomic status; ICC = intraclass correlation. Betas are unstandardized.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Table 4.

Mixed Model Regressions

Predictor	β	SE β	R2	ICC	Predictor	β	SE β	R2	ICC	Predictor	β	SE β	R2	ICC
Predicting the teacher-child       relationship from EC				.11	Predicting social competence from EC				.09	Predicting classroom participation from EC				.04
Sex	.34***	.09			Sex	.13*	.06			Sex	.13***	.04
SES	−.01	.06	.09		SES	.18***	.04	.16		SES	.05*	.02	.17
EC	.61***	.10	.21		EC	.48***	.07	.28		EC	.32***	.04	.32
Predicting spring GPA from EC and       the teacher-child relationship				.14	Predicting spring GPA from EC and       social competence				.14	Predicting spring GPA from EC and       classroom participation				.14
Fall GPA	.52***	.06			Fall GPA	.51***	.06			Fall GPA	.45***	.06
Sex	.10	.08			Sex	.12	.08			Sex	.08	.08
SES	.11*	.06	.47		SES	.08	.06	.47		SES	.09	.05	.47
EC	.45***	.10	.52		EC	.44***	.10	.52		EC	.38***	.10	.51
Teacher-child       relationship	.12*	.06	.53		Social       competence	.20*	.08	.53		Classroom       participation	.66***	.14	.56
Predicting spring absence from EC and       the teacher-child relationship				.03	Predicting spring absences from EC       and social competence				.02	Predicting spring absences from EC and       classroom participation				.02
Fall absence	.41***	.06			Fall absences	.42***	.06			Fall absences	.41***	.06
Sex	.06	.09			Sex	.01	.09			Sex	.08	.09
SES	.11†	.06	.19		SES	.12	.06	.19		SES	.14*	.06	.19
EC	−.15	.10	.21		EC	−.23*	.11	.21		EC	−.06	.11	.21
Teacher-child       relationship	−.15**	.06	.23		Social       competence	−.01	.08	.21		Classroom       participation	−.55***	.15	.25

Note. EC = children’s effortful control; SES = family socioeconomic status; ICC = intraclass correlation. Betas are unstandardized. Sex was dummy coded such that boys = 1 and girls = 2.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

The variables did not exceed West, Finch, and Curran’s (1995) cutoffs for skewness, kurtosis, and outliers. According to Cook’s (1977) distance, there were no multivariate outliers. In each analysis, we controlled for children’s sex and family SES. In addition, when predicting spring GPA (or spring absences), we controlled for fall GPA (or fall absences) to examine the relation of the predictors to change in academic competence across the school year.

To address the first goal of predicting indices of academic competence from EC, we computed two mixed model regressions (see Table 3). Consistent with expectations, EC was positively related to spring GPA and negatively related to spring absences. To test the prediction that the teacher–child relationship, social competence, and classroom participation partially mediated the relation between EC and GPA (or absences), we computed additional mixed model regressions on the basis of the guidelines outlined by MacKinnon, Lockwood, Hoffman, West, and Sheets (2002). For mediation to be present, EC should predict the mediator (e.g., teacher–child relationship, social competence, or classroom participation) and the mediator should predict the outcome (e.g., GPA or absences) when EC is included in the model. Full mediation exists if the relation between EC and the outcome is zero when the mediator is included in the model, whereas partial mediation exists if EC continues to predict the outcome when the mediator is in the model.

The first column of betas in Table 4 illustrates that the relation between EC and the teacher–child relationship was significant and that the teacher–child relationship was positively related to GPA and absences beyond the effects of EC (and fall GPA or absences, sex, and SES). To test for mediation and to accommodate the nonnormal distribution of the indirect effects (e.g., the product of the coefficient from the independent variable to the mediator and the coefficient from the mediator to the dependent variable are generally nonnormally distributed), we used a confidence interval method (MacKinnon et al., 2002). The upper and lower confidence limits are based on the product of the two random variables from tables produced by Meeker, Cornwell, and Aroian (1981). When the confidence intervals do not include zero, mediation is significant. As shown in Table 5, the confidence limits for mediation by the teacher–child relationship of the EC to academic competence (GPA and absences) relation do not include zero. Table 5 also indicates that the teacher–child relationship mediated 14% of the effect of EC on GPA and 40% of the effect of EC on absences.

Table 5.

Confidence Limits for Mediated Effects

Effect	Lower confidence limit	Upper confidence limit	Total effect that was mediated
EC > TCR > GPA	.01	.15	14%
EC > TCR > Absence	−.18	−.02	40%
EC > SC > GPA	.02	.18	19%
EC > SC > Absence	−.08	.07	1%
EC > PART > GPA	.12	.32	36%
EC > PART > Absence	−.29	−.08	76%

Note. EC = children’s effortful control; TCR = teacher–child relationship; GPA = grade point average; SC = social competence; PART = classroom participation. Bolded effects are significant at the .05 level because the confidence limits do not contain zero.

Findings in the second panel of Table 4 demonstrate that EC was positively related to social competence and that social competence partially mediated the relation between EC and GPA but not absences (see Table 5 for the confidence limits and the percentage of variance that was mediated). In contrast, consistent with the findings for the teacher–child relationship, there was evidence that classroom participation partially mediated the relation between EC and GPA and mediated the relations between EC and absences (see the third column of betas in Table 4).

Although we did not expect findings to differ by ethnicity, we computed interactions to test if the strength of the relations differed for Mexican American versus European American students, because 47% of the sample was Mexican American and the next largest percentage of ethnic population represented was European American (30%). We added the main effect of ethnicity (i.e., Mexican American vs. European American) and the interaction of ethnicity and EC when predicting the teacher–child relationship, social competence, and classroom participation. When predicting GPA or absences, we also tested if the mediators interacted with ethnicity. None of the interaction terms or main effects of ethnicity were significant. The relatively small sample sizes for the other ethnic groups (e.g., African American students) precluded the testing of further ethnic group comparisons. Thus, consistent with expectations, there was no evidence that the strength of the findings differed across the ethnicities we tested.

Discussion

Our primary purpose in this study was to begin to fill a gap in the literature on the regulatory and social variables related to children’s school success. Despite a recent research emphasis on the importance of academic functioning for positive developmental trajectories (Ladd, 2003; Welsh et al., 2001), 15% of U.S. students drop out of formal schooling prior to graduating from high school (Stoops, 2004). Students who perform poorly often develop negative attitudes and poor scholastic habits early in their school careers; thus, a better understanding of the regulatory and social factors that are related to academic achievement in early academic grades may inform intervention programs for these students. Therefore, our first goal was to test if EC was positively related to changes in children’s grades and absences, and the second goal was to test if part of the relation between EC and academic competence was mediated by children’s social relationships and classroom participation.

Results supported the hypothesis that EC was positively related to grades and negatively related to absences. This relation is consistent with limited theory and data (Eisenberg et al., 2005; Hill & Craft, 2003; Raver, 2002; Valiente et al., 2007). Blair (2002) has argued that EC, and particularly attentional regulation, is related to academic competence because students who have difficulty directing their attention and behavior likely experience significant challenges when trying to learn and focus on educational material. This explanation has roots in the cognitive literature (Ruff & Rothbart, 1996) and is supported by findings that children who have difficulty with attention often have poor reading and language skills (McGee, Partridge, Williams, & Silva, 1991; Tamis-LeMonda & Bornstein, 1989), perhaps because frequently shifting attention and moving between tasks interfere with both learning and completing tasks.

Consistent with our hypotheses, there was evidence that some of the relation between EC and grades (as well as absences) was mediated by the teacher–child relationship, social competence (for grades only), and classroom participation. To our knowledge, this is the first study to find that the teacher–child relationship and social competence mediate part of the relation between EC and grades. A number of investigators have found links between children’s EC and their social competence, problem behaviors, and social skills (see Rothbart & Bates, 2006, for a review), and data from this study suggest that part of the reason children high in EC perform well in school is indirect and through their social relationships at school. These results support the hypothesis that students’ relationships in the school context are important for school success (Baumeister & Leary, 1995; Furrer & Skinner, 2003). It is also possible that when children are low in EC and disruptive in class, they receive less classroom support from teachers and peers, miss out on learning opportunities, and view the classroom environment negatively and as something to be avoided.

In addition to the teacher–child relationship and social competence, classroom participation partially mediated the relation between EC and GPA and mediated the relation between EC and absences. These results are consistent with findings that classroom participation is positively related to math and language skills (Buhs & Ladd, 2001) and improvements in academic achievement (Ladd, Buhs, & Seid, 2000) and with evidence that school liking mediates the relation between EC and academic competence (Valiente et al., 2007). The data presented here coincide with Wentzel’s (1999) hypothesis that children who are comfortable and engaged at school may also perform well academically because they are motivated to pursue goals valued in the school context.

There is increasing evidence that indices of EC are related to students’ academic competence. In this article, we have argued that part of the reason for this relation is that EC provides students with both relational and motivational advantages that help them perform well. Because we found evidence of partial mediation (especially for grades), there is some evidence that EC also has a direct effect on academic competence. Perhaps there are components of EC that are more closely tied to academic performance and that are independent of relational and motivational processes. In future work, it would be useful to measure various components of EC and related constructs to more closely assess why there are both direct and indirect effects. More cognitively oriented components of EC such as planning and attention allocation may be directly related to academic competence. Inhibitory components are necessary for desirable behavior, and these may be aspects of EC that are mediated by constructs such as social competence, the teacher–child relationship, and classroom participation. One could test the working hypothesis that social and motivational processes mediate the relational and inhibitory components of EC, but the attentional advantages directly relate to academic competence by obtaining purer measures of the components of EC. Advancing this approach is one way to more fully explain why preschoolers’ ability to delay gratification is associated with later verbal and quantitative SAT scores (Shoda et al., 1990).

This study demonstrates several strengths. First, we incorporated data from multiple reporters for all variables (i.e., parents, teachers, and children reported on the same variables) to reduce common source variance. Second, although researchers have recently attended to the influence of EC on academic competence, few have examined processes or mechanisms underlying this relation. A more precise understanding of why relations emerge between EC and academic performance and school absence is useful for promoting children’s positive development. Third, the inclusion of a large percentage of Mexican American participants strengthens this study: The U.S. Bureau of the Census (2004) reported that by 2050, 25% of students will be of Latino descent, yet research on the normative academic functioning of this population is rare. Understanding school success among ethnic minority students is particularly important because a robust association between ethnic minority status and the likelihood of failing to complete high school has been established, with Latino students ranked most likely to drop out (Kaufman, Alt, & Chapman, 2004; U.S. Bureau of the Census, 2004). In this study, results did not differ for the European American participants and the Mexican American participants.

Finally, although not all variables were assessed longitudinally, academic competence was examined at two time points. Prediction of children’s academic competence in the spring was examined while controlling for their academic competence in the fall. By controlling for fall grades or absences when examining regulatory and social contributors to children’s spring grades or absences, one can assess how these factors related to academic competence beyond children’s preexisting academic ability.

Despite strengths, this study had some limitations. First, we used concurrent assessments of children’s EC, social competence, and classroom participation. Our data are thus correlational and do not allow for firm conclusions about directionality. Second, although the scores used in this study are valid and relate to observed indices of the relevant constructs (Birch & Ladd, 1997; Capaldi & Rothbart, 1992; Eisenberg et al., 2000), the data in the current study were assessed with questionnaires only, and future studies would benefit from using observational assessments. For example, Kochanska and colleagues have developed a battery of tasks to measure young children’s EC, and methods are available to observationally code both student-teacher interactions and engagement (Kochanska et al., 2000, 2001; Ladd et al., 1999; Pianta, La Paro, Payne, Cox, & Bradley, 2002). These methods will be especially useful tools to elaborate on the beginning stages from this line of research. Finally, this line of research would be strengthened by adding a measure of IQ to the models. Evidence indicates that measures of children’s regulatory abilities, relationships, and participation relate to academic competence beyond the effects of IQ (Blair, 2001; Gottfried, 1990; Masten et al., 2005), but it remains possible that the strength of the relations would be reduced after including IQ.

Despite these limitations, the findings from the current study advance the understanding of the relations between regulatory and social variables and academic competence and provide new information about mechanisms that may explain why children’s regulatory abilities are associated with their learning and school success. The results presented here provide evidence that EC is related to academic competence, that the teacher–child relationship and classroom participation partially mediate the relation of EC to GPA and absences, and that social competence partially mediates the relation of EC to GPA. These models present some possible process mechanisms underlying factors that are associated with children’s academic competence, and these findings emphasize the importance of considering regulatory and social influences on academic competence in future models.