Self-Regulation in Early and Middle Childhood as a Precursor to Social Adjustment Among Low-Income, Ethnic Minority Children

Abstract

Although existing research has shed much light on the development of ethnic minority children, many studies focus on maladjustment, such as behavioral problems, without also speaking to positive experiences in children’s lives, such as friendship. An aspect of development that predicts both positive and negative outcomes for children is self-regulation. The present study investigates precursors and sequelae of self-regulation in middle childhood among low-income, ethnic minority children. The four self-regulatory constructs examined in the current study include low-level executive function (EF; e.g., working memory), high-level EF (e.g., planning), effortful control (EC; e.g., delay of gratification), and impulsivity (e.g., does not think before doing). EC in preschool was related to high-level EF and impulsivity in elementary school. High-level EF explained positive and negative aspects of social development during middle childhood. Additionally, self-regulation during elementary school played a mediating role between EC in preschool and social development in middle childhood.

Nearly 8% of low-income children experience challenges to their mental health, and the gap between low-income and high-income children experiencing such difficulties has increased (Aratani, Stagman, Brellochs, & Tilipman, 2012). Still, there is heterogeneity in development among low-income children, where adjustment falls along a continuum. Indeed, developmental scholars have advanced our understanding of resilience (Rutter, 2013), where some low-income children thrive despite facing economic disadvantage. Self-regulation, a broad term that encompasses children’s ability to manage their own behavior, attention, and emotions (Nigg, 2017), may serve as a source of resilience for low-income children’s broader development in the context of poverty (Rutter, 2013).

Notably, below the poverty line, there is a disproportionate representation of ethnic minority children (Jiang, Granja, & Koball, 2017), whose development is still often viewed from a deficit perspective (Cabrera & Leyendecker, 2017). Scholars continue to call for a greater focus on positive development among minority children, where studies examine factors related not only to lower behavioral problems but also to better social adjustment (Cabrera & Leyendecker, 2017). As such, the current study aims to shed light on positive and negative aspects of social development among low-income, ethnic minority children during middle childhood, with a focus on the key developmental process of self-regulation.

More specifically, the present investigation is guided by the following goals. First, given the importance of early childhood, we examined the ways self-regulation during early childhood predicts self-regulation during middle childhood. Second, we investigated self-regulation during middle childhood as a precursor to both positive and negative outcomes in the domain of social development, which includes friendship quality, engagement in promotive activities, and behavioral problems. Lastly, we tested middle-childhood self-regulation as a mediator of the link between early childhood self-regulation and middle-childhood social adjustment.

Self-Regulation

Following Nigg’s (2017) recent review of the existing self-regulation literature, we define self-regulation as broadly encompassing both top-down and bottom-up processes that modify cognition, behavior, and emotion. Furthermore, we use Nigg’s (2017) framework in our conceptualization of the following specific self-regulation constructs: executive function, effortful control, and impulsivity. Executive function (EF) refers to top-down cognitive functions that play a role in the control of behavior, emotion, and cognition (Diamond, 2013; Nigg, 2017; Willoughby & Blair, 2011). This involves both lower-level EF, with components such as working memory (e.g., combination of short-term memory and focused attention), which occurs within a short temporal window (e.g., minutes), and higher-level EF, such as planning, which occurs within a longer time frame (e.g., hours or longer; Nigg, 2017).

Related but distinct from EF is effortful control (EC). Operating at the dispositional trait level, EC reflects the tendency to be able to draw on cognitive control (top-down processes) to regulate one’s automatic responses (e.g., bottom-up processes such as emotional reactivity; Eisenberg, Smith, & Spinrad, 2011; Rothbart, Sheese, & Posner, 2007). EC has been captured by using delay-of-gratification tasks that involve children being asked, for instance, to withhold from reaching toward a desirable gift (Murray & Kochanska, 2002). Like lower-level EF, the temporal focus for EC (e.g., waiting 1 minute to open a gift) is more immediate. In the current study, lower-level EF and EC were used to reflect self-regulation during the preschool years.

Impulsivity

Impulsivity is a nonreflective action that may involve bottom-up processes (e.g., spontaneous reaction to a desirable reward) and top-down processes (e.g., the ability to replace one response with another, regardless of the context; Nigg, 2017). Like EC, impulsivity involves both bottom-up and top-down processes. In contrast to EC, but like EF, impulsivity can involve a longer time frame (e.g., choosing an immediate reward over a delayed reward with a wait time of an hour or longer). High-level EF and impulsivity measures were used to tap self-regulation during middle childhood in the present investigation. Notably, the self-regulatory constructs examined here map onto the conceptualization of “hot” and “cool” EF, where hot EF is used in situations that are emotionally motivated, but cool EF is employed in purely cognitive scenarios (Zelazo, 2015). Within this framework, impulsivity and EC fall under the hot EF category, and low-level and high-level EF would be considered forms of cool EF.

Links Across Self-Regulation Constructs During Early and Middle Childhood

To our knowledge, the existing literature on children’s self-regulation tends to focus on self-regulatory processes within early childhood rather than across early and middle childhood. Previous research has rarely tested for longitudinal relations across different types of self-regulation during early and middle childhood (Harms, Zayas, Meltzoff, & Carlson, 2014; Raver, McCoy, & Lowenstein, 2013), although this would help clarify the use of different types of self-regulatory terms (Jones, Bailey, Barnes, & Partee, 2016; Morrison & Grammer, 2016; Nigg, 2017). This gap in prior research can be attributed to the study of EF and EC tending to occur in separate literatures (Diamond, 2013; Nigg, 2017; Rothbart et al., 2007). Links among various self-regulation constructs over time may take multiple forms. Morrison and Grammer (2016) state that there could be three types of linkages across different aspects of self-regulation over time. First, different types of self-regulation could develop in parallel to one another. Second, early EF may help prevent later impulsive behavior in the classroom. Third, self-regulation developed in preschool may aid the formation of later EF.

Some support exists for the first path outlined by Morrison and Grammer (2016). Longitudinal links have been found between EF assessments across early childhood (Raver et al., 2013) and across early and middle childhood (Friedman et al., 2014). Still, largely absent from the extant literature is investigation of associations from lower-level EF in preschool to higher-level EF during middle childhood in the context of more rigorous models that include multiple measures of self-regulation, control for a wide range of demographic characteristics, and are estimated among low-income, ethnic minority children (Jones et al., 2016; Morrison & Grammer). This is a notable gap to fill given that demographic factors such as having lower household income have been identified as risk factors for lower self-regulation (Raver et al., 2013).

Regarding the second and third paths described by Morrison and Grammer (2016), there has not been extensive study on the degree to which low-level EF during preschool may predict less impulsivity in middle childhood, nor on the extent to which EC in preschool may predict high-level EF during middle childhood. On one hand, it could be that EF emerges in early developmental stages, manifesting first as infants’ growing ability to focus attention, and once EF develops more fully, it reduces the likelihood of displaying impulsivity in elementary school (Brocki, Eninger, Thorell, & Bohlin, 2010).

On the other hand, it could be that preschoolers’ EC enables higher-level EF. At younger ages, children are likely to engage in self-regulatory strategies that are more behavioral in nature, such as closing their eyes to ignore stimuli (e.g., a desirable gift), instead of using approaches that involve more cognitive processes (e.g., pretending that a desired treat is unappealing; Mischel, Shoda, & Rodriguez, 1989). Over time, children are better able at engaging in cognitive strategies as self-regulation becomes more internalized. During middle childhood, children are better equipped to use metacognition, which enables them to reflect on their own thoughts and to employ strategies such as planning, self-monitoring, and self-correction (Friedman et al., 2014). In fact, a prior study predicted EF at age 18 from a self-regulation composite that included EC in middle childhood (Comas, Valentino, & Borkowski, 2014).

Links From Self-Regulation to Broader Social Development

In addition to studying predictors of middle-childhood self-regulation, we examined middle-childhood self-regulation as a predictor itself. We aim to expand knowledge on the benefits of self-regulation by investigating high-level EF and impulsivity in middle childhood as predictors of three aspects of social development. First, friendship deserves attention given that children begin spending considerably more time with peers during middle childhood (Brown, Dietz, Rubin, Bukowski, & Laursen, 2009). Higher EC has been linked to more popularity, which is often based on the number of children in a classroom who report that they like each classmate (Eisenberg et al., 2011). Yet, to our knowledge, existing research has not extensively examined the degree to which quality of friendship may be predicted by higher-level EF or impulsivity. Here, we take a first step at examining these links by capturing friendship quality specifically in terms of caring behavior between friends, which involves peers validating other children’s skills, ideas, and worth (Parker & Asher, 1993). Better friendship quality may be particularly meaningful for low-income, ethnic minority children given that exposure to poverty-related stressors can strain the quality of relationships (Lengua et al., 2014).

Second, self-regulation in middle childhood may explain the number of promotive factors that children experience. Promotive factors have been conceptualized as individual or contextual factors that increase the chance of children’s better social adjustment (Zimmerman et al., 2013). Participation in both after-school programs and volunteer activities has been positively linked to well-being (Mahoney, Parente, & Lord, 2007; Scales, Blyth, Berkas, & Kielsmeier, 2000). As such, we conceptualize participation in after-school programs (e.g., enrichment activities related to sports and the arts) as a school-related promotive factor (Mahoney et al., 2007; Zimmerman et al., 2013), and participation in volunteer activities (e.g., older children helping others in their communities) as a community-related promotive factor (Eisenberg, Morris, McDaniel, & Spinrad, 2009; Zimmerman et al., 2013).

School-age children’s high-level EF and impulsivity may predict how many school-related and community-related activities fill their lives. Impulsivity may interfere with engagement in multiple promotive activities, but high-level EF may enable children to shift attention toward and to maintain focus on multiple promotive activities, despite competing demands. As such, high-level EF and lower impulsivity may not only predict a wide range of positive outcomes in the social domain (Diamond, 2013; Martel, Levinson, Lee, & Smith, 2017), but also the number of promotive factors children may experience, which can be captured by cumulative indices (Jennings et al., 2016; Zimmerman et al., 2013). Yet, we lack knowledge on the prediction of cumulative promotive factors from high-level EF and impulsivity. This gap in existing research is worth addressing given that low-income, ethnic minority children are likely to face cumulative risk factors (Lengua et al., 2014).

Third, taking a comprehensive approach, we also study the extent to which high-level EF and lower impulsivity may contribute to fewer behavior problems. Previous research has documented links from greater EF and less impulsivity to fewer behavior problems (Diamond, 2013; Eisenberg et al., 2011; Han et al., 2016; Moran, Lengua, & Zalewski, 2013; Morris et al., 2013; Motamedi, Bierman, & Huang-Pollock, 2016; Rudolph, Troop-Gordon, & Llewellyn, 2013; Sasser, Bierman, & Heinrichs, 2015; Sulik et al., 2015). However, we know less about specific associations from different types of self-regulation to particular aspects of behavior problems.

Research on links among particular dimensions of self-regulation and subscales of problematic behavior offers a more precise lens on child development, which is particularly needed among older children whose behavior problems tend to be more differentiated than those of younger children (Campbell, 2006). Thus, we focused on four problem-behavior subscales that have not been extensively studied as outcomes in models with different types of self-regulation as predictors. These four dimensions of behavior problems include anxiety (i.e., fear and nervousness; Cooper-Vince, Pincus, & Comer, 2014; Eisenberg et al., 2001), social withdrawal (i.e., being shy and preferring to be alone; Cole, Zapp, Fettig, & Perez-Edgar, 2016; Eisenberg et al., 2001), hyperactivity (i.e., difficulty with sitting still; Martel & Nigg, 2006), and inattention (i.e., struggles to concentrate and to follow directions; Martel & Nigg, 2006). Notably, anxiety and social withdrawal have been studied as internalizing problems (Eisenberg et al., 2001), and hyperactivity and inattention have been examined as externalizing behaviors (Nakamura, Ebesutani, Bernstein, & Chorpita, 2009).

Self-Regulation in Middle Childhood as a Mediator

Lastly, we tested whether high-level EF and impulsivity underlie the link from EC and lower-level EF in preschool to social adjustment in middle childhood, as part of our effort to advance our understanding of the ways that EC and lower-level EF benefit children’s broader development. The impact of preschoolers’ EC and EF on many social outcomes has been studied (Diamond, 2013; Eisenberg et al., 2011). However, we lack research on the factors that explain such linkages. One exception involves Sasser and colleagues (2015) who found that prekindergarten children’s learning-related behavior, which included self-regulation and attentional focus, mediated the link from EF in preschool to their social adjustment in third grade, but that past study did not examine EC in particular. Building on a foundation of EC in preschool, high-level EF may be particularly beneficial in middle childhood as children’s social settings widen, their peer relationships deepen, and school-age children engage in more social information processing (Holmes, Kim-Spoon, & Deater-Deckard, 2016; Kochanska, 1994). Still, to our knowledge, there has not been widespread research on the ways in which impulsivity vs. high-level EF may mediate associations from lower-level EF and EC in preschool to social adjustment in school-age children. Such an overall picture would help fill gaps in what is known regarding the roles of different aspects of self-regulation in broader social and emotional well-being during middle childhood (Jones et al., 2016; Morrison & Grammer, 2016; Nigg, 2017).

Research Aims

Guided by the extant literature, the current investigation pursued the following research objectives. First, we assessed the degree to which EC and lower level EF in preschool uniquely explained their impulsivity and high-level EF during middle childhood. Second, we examined school-age children’s impulsivity and high-level EF as distinct predictors of their later friendship quality, cumulative promotive factors, and behavioral problems. Third, we examined how self-regulation in preschool and middle childhood may play roles in mediating pathways. Specifically, high-level EF and impulsivity in middle childhood were tested as unique mediators of linkages from EC and low-level EF in preschool to friendship quality, cumulative promotive factors, and behavioral problems in elementary school. Notably, we also sought to make a contribution to the literature by testing rigorous models that included a host of demographic characteristics and three waves of data (Jones et al., 2016).

Method

Participants and Procedure

Children were participants in the Chicago School Readiness Project (CSRP), which originally included 602 children who attended 18 Head Start programs in seven low-income communities. The CSRP was a classroombased, multicomponent mental health intervention, which was administered within the context of a cluster randomized controlled trial (RCT) design (for details, see Raver et al., 2011). Table 1 provides descriptive statistics, which show that a substantial percentage of families faced risk. Indeed, 85% of the sample fell below 200% of the poverty threshold, which is used to determine low-income status (Jiang et al., 2017). Analyses include data from the intervention year, which we refer to as the baseline and preschool year. Children and their families and teachers participated in additional waves, and the current study includes data from 4 and 6 years after baseline. The three waves of data are referred to as T1, T2, and T3. Notably, results from attrition analyses suggest that children who were retained in the sample vs. not were generally similar, except that a smaller proportion of Hispanic children were retained (24% vs. 38%) and a larger percentage of African American children were retained (69% vs. 53%). Still, there were no significant differences in other demographic factors such as income, effortful control, and low-level executive function at baseline (T1).

Table 1.

Descriptive statistics on demographic characteristics, self-regulation, friendship quality, promotive factors, and behavioral problems

	M/%	SD	Min.	Max.	N
Child characteristics
Age at baseline (years)	4.10	0.61	3	6	492
Age at T2 (years)	8.02	0.72	6	9	446
Age at T3 (years)	9.88	0.70	8	12	478
Male	0.47		0.00	1.00	231
Female	0.53		0.00	1.00	261
Race/ethnicity
Latino	0.24		0.00	1.00	492
African American	0.69		0.00	1.00	492
Other	0.07		0.00	1.00	492
Family characteristics
Parent single	0.70		0.00	1.00	468
Income-to-needs ratio	0.68	0.61	0.00	3.57	438
Maternal educational risk	0.25		0.00	1.00	448
Maternal employment risk	0.42		0.00	1.00	484
Self-regulation (baseline)
Effortful control	0.02	0.84	−3.53	3.64	413
Low-level EF	0.00	0.67	−2.63	0.69	414
Self-regulation (T2)
High-level EF	0.38	0.26	0.00	1.00	387
Impulsivity	0.31	0.27	0.00	1.00	388
Outcomes (T3)
Friendship quality (C)	3.98	0.84	1.00	5.00	387
Promotive factors (P)	3.79	1.77	0.00	7.00	480
Promotive factors (T)	1.32	1.22	0.00	4.00	332
Anxiety (T)	0.15	0.19	0.00	1.06	360
Social withdrawal (T)	0.19	0.28	0.00	1.63	358
Hyperactivity (T)	0.42	0.48	0.00	2.00	359
Inattentive (T)	0.52	0.50	0.00	1.86	358

Note. For low-level executive function (EF) at T1, higher scores indicate more EF. However, high-level EF at T2 is coded in a negative direction such that greater scores indicate lower levels of EF, consistent with the original measure (McCoy et al., 2011). T2 = Time 2; T3 = Time 3; T = teacher report; P = parent report; C = child report.

Multiple sources of data were used, including direct assessment, and child-, parent-, and teacher-reported measures. Direct assessments of children’s EC and low-level EF were collected during the fall of the preschool year. At T2, teachers reported on children’s impulsivity and high-level EF. Outcomes were captured at T3, where children reported on their friendship quality, both teachers and parents reported on children’s cumulative promotive factors, and teachers reported on children’s behavioral problems. Parents reported on child and family background characteristics during the baseline year. Written informed consent was obtained from all participants prior to their initial participation in CSRP and was acquired for each subsequent wave. The procedure used for obtaining consent was institutional review board approved and included standard components such as a description of the project, its purpose, the voluntary nature of participation, and issues pertaining to confidentiality.

Measures

Self-regulation.

Four measures of self-regulation were employed, including two assessments at baseline (T1) and two measures at T2. At T1, direct assessments of self-regulation in preschool tapped EC and lower-level EF by using the Preschool Self-Regulation Assessment (PSRA; Smith-Donald, Raver, Hayes, & Richardson, 2007). Data on EC were yielded from four tasks developed by the work of Kochanska and colleagues: the toy wrap, toy wait, snack delay, and tongue tasks (see Murray & Kochanska, 2002). Performance scores on these tasks were standardized and then averaged into an effortful control score (α = 0.77). Lower-level EF was captured by using two tasks, the balance beam task (Murray & Kochanska, 2002) and the pencil tap task, which was adapted from the pegtapping task (Blair, 2002; Diamond & Taylor, 1996). Scores on these tasks were standardized and then averaged (α = 0.86). Together, these two tasks reflect EF more broadly (Raver et al., 2011; Smith-Donald et al., 2007). Aligned with Nigg (2017), we refer to our preschool measure of EF as low-level EF given the lower cognitive demand of the balance beam and pencil tap tasks (e.g., children use working memory to hold task instructions in their minds).

Teachers rated children’s self-regulation at T2 (McCoy, Raver, Lowenstein, & Tirado-Strayer, 2011). In particular, McCoy and colleagues (2011) sought to create a validated teacher-report measure that captured multiple domains of school-age children’s self-regulation and could easily be administered in schools serving children at risk. Using items from the Barratt Impulsiveness Scale Version 11 (BIS-11; Patton, Stanford, & Barratt, 1995) and the Behavior Rating Inventory of Executive Function (BRIEF; Gioia, Isquith, Guy, & Kenworthy, 2000), McCoy et al. conducted both exploratory and confirmatory factor analyses. A two-factor structure emerged, where one factor reflected behavior dysregulation and the other reflected cognitive dysregulation. Tests of measurement invariance suggested that this factor structure held across subgroups based on race, gender, and high vs. low poverty risk. High Cronbach’s alphas were found for the cognitive and behavior aggregates in each of these subgroups, and tests of criterion validity yielded significant bivariate correlations between these two composites and other measures of behavior (see McCoy et al. for details).

More specifically, questions from the BIS-11 reflected attention, activity, and planning (e.g., says things without thinking) by using a scale of 1 (rarely/never) to 4 (almost always/always). Based on a metric from 1 (never) to 3 (often), items from the BRIEF tapped working memory and inhibitory control. Items were standardized and aggregated into a composite originally referred to as behavioral dysregulation (α = 0.97; McCoy et al., 2011), which used 7 items from the BIS-11 and 10 items from the BRIEF (e.g., impulsive). A composite originally labeled as cognitive dysregulation (α = 0.96; McCoy et al., 2011) employed 9 items from the BIS-11 and 10 items from the BRIEF (e.g., planning). In an effort to map the constructs examined here with recent efforts in the literature to clarify self-regulation terminology (Nigg, 2017), the current study refers to the former factor as impulsivity and the latter factor as high-level EF. Notably, the direction of the scores is consistent with the original labels, meaning that higher scores on impulsivity reflect less self-regulation, and lower scores on high-level EF indicate more self-regulation.

Friendship quality.

As a step toward better understanding low-income children’s friendship quality in the context of a larger survey study at T3, we asked children to rate six statements such as “my best friend would always stick up for me” and “my best friend and I make each other feel important and special.” Five of the six items were from the validation and caring subscale of the Friendship Quality Questionnaire (Parker & Asher, 1993). These included the highest loading behaviorally oriented items on that subscale, where the highest loadings ranged .66–.78. We also included a sixth behaviorally oriented item (“my best friend and I look out for each other”). All six items were rated on a scale of 1 (not at all true) to 5 (really true), and an average of the six items was used to form a composite score (α = 0.77).

Cumulative promotive factors.

To provide a broad window into low-income children’s exposure to promotive factors at T3, we used reports from both parents and teachers. Past research has found children’s positive development to be linked to their participation in after-school programs (Mahoney et al., 2007), volunteer service (Eisenberg et al., 2009), civic engagement (Guillaume, Jagers, & Rivas-Drake, 2015), and care of a family member (Tsai, Telzer, Gonzales, & Fuligni, 2015). Given parents’ wider understanding of children’s lives, parental reports of promotive factors reflected a range of experiences, such as whether children volunteered, participated in an after-school club, were a member on a sports team, took care of a family member, accompanied an adult to vote, participated in a band or performance group, or were placed on the honor roll. In contrast, given teachers’ more in-depth knowledge of children’s school experiences, teacher reports focused on school-related promotive factors, which included whether students took part in an after-school program, were a member of a sports team, participated in a band or performance group, or were listed on the honor roll. Participation in each of the activities was coded as a 1, and no participation in each activity was coded as a 0. In line with existing research on cumulative promotive factors (Jennings et al., 2016; Zimmerman et al., 2013), scores on these dummy variables were summed, which enabled us to capture promotive factors parsimoniously. The parent-reported cumulative promotive factor index ranged 0–7 and a teacher-reported cumulative promotive factor index ranged 0–4.

Behavior problems.

Using the Child Behavior Checklist (Achenbach, 1991), a widely used, well-established measure of behavioral problems, teachers rated children’s behavior at T3. Based on a metric ranging from 0 (not true) to 2 (very true or often true), teachers rated how true each item was at the time or within the past 2 months. Items were summed to form aggregates for the following subscales: anxiety (16 items), social withdrawal symptoms (8 items), hyperactivity (12 items), and inattention (14 items).

Background characteristics.

Given that child and family factors may be related to children’s self-regulation and broader social adjustment (e.g., Raver et al., 2013), models took into account the following background characteristics at baseline (T1). Children’s age was measured in months, gender was coded as 1 for male, and race and ethnicity were captured by using three categories: Latino, African American (omitted group), and White or other race and ethnic group. Regarding family characteristics, we included whether households were led by a single parent and families’ income-to-needs ratio, which was the household income divided by the poverty threshold. Also, we took into account mothers’ educational risk (1 = less than a high school diploma or equivalent degree [i.e., GED] and 0 = high school diploma/GED or greater) and employment risk (1 = worked less than 10 hr per week and 0 = worked 10 hr or more per week). Finally, we controlled for whether sites were randomly assigned to treatment status (1 = treatment), and whether children were in the first or second intervention cohort (1 = first cohort).

Analytic Plan

Our analytic steps included estimating correlation coefficients (see Table 2) and then focused on addressing our main research questions by using seven separate models that predicted each of the following outcomes at T3: friendship quality, parent reports of promotive factors, teacher reports of promotive factors, anxiety, social withdrawal, hyperactivity, and inattention. In each model, low-level EF and EC at T1 were tested as predictors, and impulsivity and high-level EF at T2 were estimated as mediators of the relation between self-regulation in preschool at T1 and outcomes during elementary school at T3 (see Figure 1). As such, the unique contributions of low-level EF, EC, impulsivity, and high-level EF were assessed. Moreover, the mediation models also included the following covariates: children’s age, gender, and race/ethnicity, family structure, household income-to-needs ratio, mothers’ education and employment risk, treatment status, and cohort.

Table 2.

Correlations

	1	2	3	4	5	6	7	8	9	10	11
1. Age at baseline	—
2. Effortful control	.43***	—
3. Low-level EF	.32***	.32***	—
4. High-level EF	−.02	−.19***	−.16**	—
5. Impulsivity	−.09	−.11	−.17**	.64***	—
6. Promotive factors (T)	.09	.20***	.11	−.36***	−.21***	—
7. Promotive factors (P)	.15***	.17***	.11**	−.18***	−.10**	.35***	—
8. Friendship quality (C)	.13**	.11**	.18**	−.21***	−.15**	.12**	.15**	—
9. Anxiety (T)	−.05	−.09	−.01	.20***	.11	−.14**	.00	−.07	—
10. Hyperactivity (T)	−.12**	−.14**	−.17**	.41***	.62***	−.20***	−.11**	−.12**	.31***	—
11. Inattention (T)	−.12**	−.18**	−.12**	.51***	.42***	−.40***	−.20***	−.15**	.35***	.62***	—
12. Social withdrawal (T)	−.04	−.09	−.01	.19**	−.01	−.28***	−.13**	−.14**	.46***	.11**	.48***

Note. For low-level executive function (EF) at T1, higher scores indicate more EF. However, high-level executive function at T2 is coded in a negative direction such that higher scores indicate lower levels of executive function, consistent with the original measure (McCoy et al., 2011). T = teacher report; P = parent report; C = child report.

^*p < .05.

^**p < .01.

^***p < .001.

Figure 1.

Conceptual model of associations among preschool self-regulation, middle-childhood self-regulation, and social adjustment outcomes. This figure illustrates the conceptual model guiding the current study. Direct paths of interest are bold. Nonbold paths indicate that models took into account links from preschool self-regulation to social adjustment outcomes. The four types of path a each represent the association between preschool self-regulation and middle-childhood self-regulation, and the two types of path b each reflect the link from middle-childhood self-regulation to social adjustment outcomes. For ease of interpretation, baseline covariates (i.e., age, gender, child race/ethnicity, single parent, income-to-needs ratio, education risk, employment risk, cohort, and treatment) and correlations among concurrent predictor variables are not shown.

Instead of using the traditional Baron and Kenny (1986) method of testing for mediation, we performed tests of mediation by using bootstrapped sampling to estimate mediated effects. Whereas the Baron and Kenny method detects an indirect effect by a reduction in size of the direct path after a mediator is added to a model, bootstrapping estimates an indirect-effect standard error that provides a criterion for statistical significance (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002; Preacher & Hayes, 2004). In brackets in Figures 2–4, we present the upper and lower confidence limits, which were estimated by using 5,000 bootstrapped draws from the sample. Path models were estimated by using the model indirect command in Mplus 7.0 (Muthén & Muthén, 2012), using full information maximum likelihood (FIML) estimation. FIML accommodates missing data by estimating the covariance matrix with all of the available information provided by the independent variables, effectively enabling us to use the entire sample with data on the independent variables (Enders & Bandalos, 2001).

Figure 2.

Estimates of mediation models predicting friendship quality and cumulative promotive factors. Unstandardized coefficients are presented for paths a and b. Standard errors for paths a and b are in parentheses. Unstandardized coefficients and standard errors for significant indirect effects are reported, with the lower-bound confidence interval and upper-bound confidence interval for indirect effects in brackets. P = parent report; T = teacher report; greater scores on high-level executive function indicate less executive function, which is consistent with the original measure (McCoy et al., 2011); promotive = cumulative promotive factors. * p < .05. ** p < .01. *** p < .001.

Figure 4.

Estimates of mediation models predicting hyperactivity and inattention problems. Unstandardized coefficients are presented for paths a and b. Standard errors for paths a and b are in parentheses. Unstandardized coefficients and standard errors for significant indirect effects are reported, with the lower-bound confidence interval and upper-bound confidence interval for indirect effects in brackets. Greater scores on high-level executive function indicate less executive function, which is consistent with the original measure (McCoy et al., 2011). * p < .05. ** p < .01. *** p < .001.

Results

Predicting Middle-Childhood Self-Regulation

Figures 2–4 display significant findings from mediation models estimated for each of the seven outcomes. Figure 2 focuses on friendship quality and promotive factors, Figure 3 displays statistics for anxiety and social withdrawal, and Figure 4 centers on hyperactivity and inattention. The arrows on the left represent path A, where self-regulation in preschool was estimated as a predictor of self-regulation at school age. Coefficients and standard errors for path A are displayed above each arrow on the left.

Figure 3.

Estimates of mediation models predicting anxiety and social withdrawal. Unstandardized coefficients are presented for paths a and b. Standard errors for paths a and b are in parentheses. Unstandardized coefficients and standard errors for significant indirect effects are reported, with the lower-bound confidence interval and upper-bound confidence interval for indirect effects in brackets. Greater scores on high-level executive function indicate less executive function, which is consistent with the original measure (McCoy et al., 2011). * p < .05. ** p < .01. *** p < .001.

In Figures 2 and 4, we see estimates of the link between low-level EF and high-level EF, controlling for EC. Note that the negative coefficients reported are due to high-level EF at T2 being coded in a negative direction such that higher scores indicate lower levels of EF, which is consistent with the original measure (McCoy et al., 2011). In Figure 2, this association is significant in the mediation model predicting teacher reports of promotive factors (b = −0.06, SE = 0.03, p < .05). In order to gauge effect size, we separately calculated the corresponding standardized coefficients for each link, which was −0.15 in the mediation model predicting teacher-reported promotive factors. Figure 4 shows consistent findings, where there is a significant link between these constructs when explaining inattention (b = −0.05, SE = 0.03, p < .05). The corresponding standardized coefficient for this relation was −0.13.

Figures 2–4 also show the link between EC and high-level EF when controlling for low-level EF. EC predicted high-level EF when mediation models included friendship quality as an outcome (b = −0.08, SE = 0.02, p < .001). Similarly, in mediation models that predicted parent reports (b = −0.06, SE = 0.02, p < .01) and teacher reports of promotive factors (b = −0.06, SE = 0.02, p < .001), EC explained individual differences in high-level EF. Figure 3 shows that statistics for path A were similar when anxiety (b = −0.08, SE = 0.02, p < .001) and social withdrawal (b = −0.07, SE = 0.02, p < .01) were predicted by mediation models. In Figure 4, the statistics for path A were the same in mediation models predicting hyperactivity and inattention (b = −0.07, SE = 0.02, p < .01). When converted to standardized coefficients, these path A coefficients for the relation between EC and high-level EF ranged 0.20–0.26. However, analyses did not reveal significant links between low-level EF and impulsivity.

In addition, EC was related to impulsivity, net of low-level EF. In Figures 3 and 4, the estimates for the link between effortful control and impulsivity were the same (b = −0.04, SE = 0.02, p < .05) when mediation models included anxiety, social withdrawal, hyperactivity, and inattention as outcomes. When converted to standardized coefficients, the relation between effortful control and impulsivity was −0.12 in magnitude.

Predicting Middle-Childhood Outcomes

Next, we focus on path B, where impulsivity and high-level EF at T2 were estimated as predictors of promotive factors, friendship quality, and behavioral problems at T3. Again, high-level EF at T2 was coded in a negative direction, where greater scores reflect lower levels of EF, which aligns with the original measure (McCoy et al., 2011). Figures 2–4 show estimates of high-level EF as a predictor, controlling for impulsivity. The standardized coefficients for these links were −0.19 for friendship quality (b = −0.61, SE = 0.23, p < .01), −0.14 for parent-reported cumulative promotive factors (b = −0.98, SE = 0.48, p < .05), and −0.33 for teacher-reported cumulative promotive factors (b = −1.53, SE = 0.33, p < .001). High-level EF explained anxiety and social withdrawal symptoms, as well. The standardized coefficient was 0.22 for anxiety (b = 0.16, SE = 0.05, p < .01), and it was 0.35 for social withdrawal (b = 0.37, SE = 0.10, p < .001). Also, high-level EF predicted inattention (b = 0.83, SE = 0.13, p < .001), where the standardized coefficient was 0.44.

Additionally, Figures 3 and 4 show estimates of impulsivity as a predictor, controlling for high-level EF. Impulsivity was negatively related to social withdrawal (b = −0.23, SE = 0.08, p < .01), as displayed in Figure 3. However, impulsivity was positively related to hyperactivity (b = 0.95, SE = 0.14, p < .001), as indicated in Figure 4. The standardized coefficient was −0.22 for social withdrawal and 0.53 for hyperactivity.

Indirect Effects

Lastly, when tests of indirect effects yielded significant results, we noted this in Figures 2–4. In Figure 2, EC at T1 was associated with friendship quality at T3 via high-level EF at T2 (b = 0.05, 95% CI [0.005, 0.094]). Similarly, EC at T1 was related to teacher-reported cumulative promotive factors at T3 through high-level EF at T2 (b = 0.09, 95% CI [0.017, 0.171]). The standardized coefficients for these indirect effects were 0.05 and 0.07, respectively.

Figure 3 shows that there were mediating pathways from EC at baseline to anxiety and social withdrawal at T3 through high-level EF at T2. There was an indirect effect for anxiety (b = −0.01, 95% CI [−0.022, −0.001]). Likewise, there was an indirect effect for social withdrawal (b = −0.03, 95% CI [−0.049, 0.005]). In terms of standardized coefficients, these indirect effects were −.05 and −.08, respectively. Furthermore, Figure 4 indicates that EC was related to teacher-reported hyperactivity through impulsivity (b = −0.04, 95% CI [−0.071, −0.002]). The standardized coefficient for this indirect effect was −0.06. In addition, EC was related to teacher-reported inattention via high-level EF (b = −0.06, 95% CI [−0.100, −0.019]). In terms of a standardized coefficient, this indirect effect was −0.10.

Discussion

In sum, the present investigation focuses on middle-childhood self-regulation and its precursors, sequelae, and role as a mediator among low-income, ethnic minority children. To our knowledge, this is the first investigation to examine different self-regulatory pathways across early to middle childhood as predictors of both positive and problematic behavior during middle childhood, in the context of a longitudinal design with three waves of data that includes a broad array of covariates representing low-income, ethnic minority children’s demographic characteristics. These features are worth underscoring given that demographic risk factors tend to jeopardize children’s self-regulation and behavioral problems (Dearing, McCartney, & Taylor, 2006; Raver et al., 2013). Also, scholars have called for more rigorous models in the study of different self-regulation constructs (Jones et al., 2016), and there is a need to focus more on positive development among ethnic minority children (Cabrera & Leyendecker, 2017). Additionally, our findings help bridge separate literatures on different self-regulation constructs (Nigg, 2017), and our results add to the existing literature, which has not extensively examined various self-regulation constructs during the transition from early to middle childhood.

Effortful Control as a Predictor

In the current study, our findings showed that EC in preschool was predictive of high-level EF during middle childhood. Additionally, low-level EF in preschool was associated with their high-level EF in middle childhood, and earlier EC was negatively related to impulsivity. These findings are congruent with prior research (Brocki et al., 2010; Comas et al., 2014; Diamond, 2013; Eisenberg et al., 2004).

Importantly, the link between EC in preschool and higher-level EF in elementary school supports one of the three paths outlined by Morrison and Grammer (2016). Identifying which of these specific pathways helps characterize children’s development and may help advance understanding of how and when intervention programs might strive to improve children’s adjustment (Morrison & Grammer). More experience with delaying gratification may provide young children with greater opportunity to internalize self-regulation strategies. As such, intervention programs might focus more on fostering EC during the preschool years as an avenue toward strengthening top-down self-regulation processes during middle childhood (Jones et al., 2016). That said, future research should continue to test the three paths delineated by Morrison and Grammer using other self-regulation measures in studies with children from a variety of income groups.

High-Level Executive Function and Impulsivity as Predictors

Regarding sequelae, we revealed links from high-level EF to friendship quality. It could be that EC, with its roots in temperament, may be especially salient for the number of friends that children make (i.e., popularity; Eisenberg et al., 2011), but that EF, and its role in social information processing, is particularly important for the way in which friends relate to one another (i.e., quality of friendship; Holmes et al., 2016). High-level EF may help children shift their attention from others and toward a best friend, and high-level EF may help children ignore distractions so that better-quality interactions occur with a best friend (Liew, Eisenberg, & Reiser, 2004; Moran et al., 2013).

In terms of cumulative promotive factors, we found that high-level EF was related to parent reports of promotive factors and linked to teacher reports of promotive factors. These results suggest that high-level EF is not only associated with a gamut of positive outcomes in the social domain (Diamond, 2013) but may help children accumulate a greater number of positive experiences. Identifying predictors of cumulative promotive factors is important given that cumulative promotive factors can yield benefits for children’s broader well-being (e.g., Jennings et al., 2016; Zimmerman et al., 2013). Children with high-level EF may be more likely to experience cumulative promotive factors because high-level EF may facilitate more effective social problem solving, engagement in positive interaction with others, and maintenance of relationships (Portilla, Ballard, Adler, Boyce, & Obradović, 2014). In turn, better social interaction may help children engage in numerous promotive activities.

Furthermore, high-level EF in middle childhood predicted fewer behavior problems where there were links to anxiety, relations to social withdrawal, and associations with inattention. In addition, impulsivity in middle childhood was associated with hyperactivity in a positive direction. High-level EF may halt rumination of fearful and nervous thoughts and facilitate the filtering out of distractions, leaving children to feel less anxious, to be less likely to withdraw from social interaction, and to display fewer inattention problems (Rudolph et al., 2013), whereas greater impulsivity increases the chance that children will have problems with hyperactivity (Martel et al., 2017). These findings are consistent with past research (Han et al., 2016; Martel & Nigg, 2006; Morris et al., 2013; Motamedi et al., 2016), but, unlike prior studies, the current study shows that these linkages are robust to the inclusion of other types of self-regulation as covariates. This is noteworthy given the “conceptual clutter” that has characterized the study of different self-regulation constructs in separate literatures (Morrison & Grammer, 2016; Nigg, 2017).

Interestingly, past research has noted relatively weak links between self-regulation and internalizing problems (Eisenberg et al., 2011). The results of the current investigation suggest that this may be due in part to differences in the types of self-regulation that predict the internalizing subscales of anxiety vs. social withdrawal. In the present study, high-level EF was related to less anxiety and fewer social withdrawal symptoms, but more impulsivity was related to less social withdrawal. The results for high-level EF are aligned with prior research findings on self-regulation as an overall predictor of adaptive outcomes for children (Cole et al., 2016; Diamond, 2013; Eisenberg et al., 2011). However, it could be that both lower impulsivity and greater social withdrawal are part of a larger overcontrolled regulatory system (Murray & Kochanska, 2002).

High-Level Executive Function and Impulsivity as Mediators

Lastly, we found that self-regulation in early childhood and self-regulation in middle childhood worked together to explain individual differences in social development via mediated pathways. In existing research, EC is often tested as a mediator between environmental experiences, such as parenting, and children’s broader development (Eisenberg et al., 2011), but we lack knowledge on processes that explain the association between children’s EC and their overall well-being. The current study helps to fill that gap by showing that higher-level EF mediated the path from EC in preschool to friendship quality and teacher-reported cumulative promotive factors, as well as to anxiety, social withdrawal, and inattention during middle childhood.

This supports the view that children’s social development is shaped by both temperament and cognition (Kochanska, 1994). During preschool, better EC may help children as they learn to share and to cooperate in a more formal group setting with new peers and adults for the first time. With major changes in children’s social worlds during middle childhood, such as the development of friendships and peer groups, children face an increasing demand to manage new information about academic expectations and social relationships (Brown et al., 2009). High-level EF in middle childhood may play an important role in focusing attention on academic instruction, processing social information, and halting rumination about social comparisons (Friedman et al., 2014; Holmes et al., 2016; Kochanska, 1994; Rudolph et al., 2013). It follows that educators and interventionists concerned with social and emotional adjustment during middle childhood might specifically target improvement in EC during the preschool years and high-level EF during elementary school.

Despite the contributions made by the present study, future research should address the current investigation’s limitations. First, future studies on low-income, ethnic minority children could examine a wider array of promotive factors and could conduct an in-depth investigation of friendship, including children’s number of friends, the quality of their friendships, and multiple reporters, such as parents, teachers, and children themselves. Second, self-regulation was measured by direct assessment in preschool but not in middle childhood. Future studies should include indicators of each measure at multiple waves and include more detailed self-regulation measures in order to take further steps toward clarifying conceptual clutter in the field (Morrison & Grammer, 2016). Third, our results are nonexperimental and should not be interpreted as reflecting causal relations. Fourth, future investigations should estimate whether our findings are generalizable to children from a wide economic spectrum and should test whether the results detected here are more protective for low-income children than higher-income children.

Regardless, the present investigation may still inform discussions about the relation and distinction among various types of self-regulation (Jones et al., 2016; Morrison & Grammer, 2016; Nigg, 2017; Sulik et al., 2015). For example, the current study found that EC was salient during early childhood and that high-level EF was important during middle childhood for better friendship quality, more promotive factors, and fewer behavior problems. Notably, these findings were detected while controlling for lower-level EF in preschool and impulsivity in middle childhood. These results speak to how various types of self-regulation may help children respond to differing contextual demands (Liew et al., 2004). In other words, children’s socially competent responses during preschool vs. elementary school may involve drawing upon different types of self-regulation. Future research on other types of self-regulation across early and middle childhood is needed. Expanding the literature in such directions may advance our understanding of the roots of overall social adjustment and inform efforts to help not only low-income, ethnic minority children, but children from various economic and ethnic backgrounds who each face opportunities and challenges during middle childhood.