Executive Function as a Mechanism Linking Socioeconomic Status to Internalizing and Externalizing Psychopathology in Children and Adolescents

Abstract

Introduction

The association between low socioeconomic status (SES) in childhood and increased risk for psychopathology is well established, but the mechanisms explaining this relationship are poorly understood. Here, we investigate the potential role of difficulties in executive functioning (EF) as a mechanism linking childhood and adolescent SES with externalizing and internalizing psychopathology.

Methods

We examined whether difficulties with EF mediated the association between SES and externalizing and internalizing psychopathology in two cross-sectional samples of children and adolescents (Study 1: N=94, ages 6–18, 51.1% male; Study 2: N=259, ages 8–16, 54.1% male) from diverse SES backgrounds in the United States. EF was measured through behavioral tasks and parent-reported behavioral regulation (BR).

Results

In both samples, children and adolescents from lower SES families were more likely to experience both externalizing and internalizing psychopathology than youth from more advantaged backgrounds and exhibited greater EF difficulties – they had lower performance on a task measuring inhibitory control and lower parent-rated BR. Reduced inhibitory control and BR, in turn, were associated with higher externalizing and internalizing psychopathology. In Study 1, difficulties with BR mediated the association of low-SES with both externalizing and internalizing psychopathology. In Study 2, low inhibitory control mediated the association between low-SES and externalizing psychopathology. These findings largely persisted after adjusting for exposure to violence, a form of adversity that is common in children from low-SES backgrounds.

Conclusions

These findings suggest that reduced EF may be an underlying mechanism through which low-SES confers risk for psychopathology in children and adolescents.

One in six children under the age of 18 in the United States live in poverty (Semega et al., 2019). Low socioeconomic status (SES) in childhood has been associated with atypical cognitive development, particularly in the domains of executive functioning and language (Duncan et al., 1994; Farah et al., 2006; Noble et al., 2006). These differences may contribute to well-documented disparities in academic achievement (Bradley & Corwyn, 2002; Brooks-Gunn & Duncan, 1997; Miller et al., 2013; Rosen et al., 2018) and increase risk for developing certain forms of psychopathology (Hatoum et al., 2017; McLoyd, 1998). Both externalizing and internalizing psychopathology are more common among children from low-SES households (Capistrano et al., 2016; McLaughlin, Costello, et al., 2012; Peverill et al., 2020). Here, we examine whether executive functions (EF)—including inhibition, cognitive flexibility, and behavioral regulation (BR)—are a potential mechanism explaining the association between SES and child psychopathology, including symptoms and features of a range of mental disorders reflecting the higher order dimensions of internalizing and externalizing.

SES-related differences in child psychopathology have been consistently documented across multiple domains of measurement including family income, parent education, and food insecurity (Brooks-Gunn & Duncan, 1997; McLaughlin, Green, et al., 2012; Peverill et al., 2020; Shankar et al., 2017). EF is a potential developmental mechanism, given that it is consistently associated with both SES and psychopathology in children and adolescents (Blair & Raver, 2016; Lawson et al., 2018; McLaughlin, 2016; Zelazo & Mller, 2002). EF is a set of cognitive skills involved in goal-directed behavior that is critical for adaptive functioning and academic success (Best & Miller, 2010; Blair & Diamond, 2008; Samuels et al., 2016). These processes include manipulating information in working memory, inhibiting automatic responses, and shifting between mental sets (i.e., cognitive flexibility) (Friedman & Miyake, 2017; Miyake, Friedman, Emerson, Witzki, & Howerter, 2000). The ability to engage in BR in daily life is inextricably tied to EF (Miyake & Friedman, 2012; Zhou et al., 2012) as an external manifestation of cognitive EF processes.

Multiple aspects of SES have been associated with EF in children and adolescents. Parent educational attainment is associated with performance on working memory and inhibition tasks (Farah et al., 2006; Noble et al., 2005; Sheridan et al., 2017). Family income is associated with working memory, inhibition, cognitive flexibility, and BR (Evans & Rosenbaum, 2008; Noble et al., 2005; Raver et al., 2013; Sarsour et al., 2011). This association exists across the entire SES distribution, not simply among families living in poverty (Noble et al., 2007). Previous research has not directly explored the association of food insecurity with EF.

EF difficulties have been consistently linked with externalizing psychopathology (Hatoum et al., 2017; Hobson et al., 2011). The association between low EF and higher externalizing psychopathology has been observed throughout development and across multiple domains of EF, including inhibition (Schoemaker et al., 2013; Wang et al., 2017) and cognitive flexibility (Schoemaker et al., 2013; Willcutt et al., 2005). Lower BR has been linked to increased risk for externalizing disorders in cross-sectional and longitudinal studies (Eisenberg et al., 2010; Lonigan et al., 2017). Although the association between EF and internalizing disorders is less consistent, inhibitory control in response to emotional stimuli is one domain of EF that has been linked to greater internalizing disorders, such as depression (Capistrano et al., 2016; Joormann & Gotlib, 2010). Individuals experiencing depression may demonstrate difficulties inhibiting irrelevant or negative emotional stimuli (Goeleven, De Raedt, Baert, & Koster, 2006), which may lead to emotional dysregulation, including rumination (Joormann, 2006; Joormann & Gotlib, 2010).

There is a dearth of existing research that has explored EF as a mediator between low SES and psychopathology in youth. One study most relevant to our hypothesis found that EF skills partially mediated the association between SES and aggression in youth (Fatima & Sheikh, 2014). To our knowledge, however, EF has not been explored as a mediator linking SES with other externalizing or internalizing psychopathology in children or adolescents.

We present a preliminary exploration of the hypothesis that difficulties in EF link SES with youth externalizing and internalizing psychopathology in two cross-sectional studies. We investigated to what degree two primary dimensions of EF (inhibitory control and cognitive flexibility), as well as the application of these skills in the service of BR, explain the association between multiple domains of SES and externalizing and internalizing psychopathology in two youth samples with wide variability in SES. These studies were designed to explore multiple components of SES separately, as opposed to as a composite measure. Previous literature has shown that SES composites do not necessarily capture, and may even obscure, substantial associations between SES and psychopathology (Call & Nonnemaker, 1999; Twenge & Nolen-Hoeksema, 2002). We also used multiple measures of EF to capture the diversity and complexity of EF-related processes (Friedman & Miyake, 2017). Previous research has demonstrated that task-based and self-report measures of EF capture complementary, yet distinct, components of cognitive control (Snyder et al., 2020). We expected that children from lower-SES households would have higher externalizing and internalizing symptoms and worse EF performance. Finally, we anticipated that variation in EF would mediate the relationship between SES and externalizing psychopathology in children and adolescents. Because low SES often co-occurs with other forms of adversity, such as physical and sexual abuse, domestic violence, and other types of violence exposure (McLaughlin, Greif Green, et al., 2012), it is important to determine whether SES-related variation in EF persists after controlling for exposure to violence (McLaughlin & Sheridan, 2016). Therefore, we additionally evaluated whether associations of SES persisted after adjusting for exposure to violence.

Study 1

Method

Sample

The sample for Study 1 included 94 children and adolescents (51.1% male) aged 6–18 years (M=13.57, SD=3.47). Participants described themselves as White (51.1%), Black (17%), Hispanic (13.8%), Asian (10.6%), and Other (7.4%). Youths were recruited for participation at schools, after-school and prevention programs, medical clinics, and the general community between February 2014 and February 2015. Recruitment efforts were targeted at recruiting a sample with variation in exposure to adversity. To do so, we recruited from neighborhoods with high levels of violent crime, from clinics that served a predominantly low-SES catchment area, and agencies that work with families who have been victims of violence to identify children with exposure to trauma, as well as a variety of sources in the general community (e.g., public transportation, libraries, after-school programs, coffee shops, etc.) to identify children without adversity experiences. Approximately half of children in the final sample had experienced some form of adversity and the other half were control participants, matched on age and sex to those with adversity experiences. See Table 1 for socio-demographic characteristics. The Institutional Review Board (IRB) approved all procedures. Informed consent was obtained from all participants in the study.

Table 1.

Descriptive Statistics of SES, Violence Exposure, Executive Function, and Internalizing and Externalizing Psychopathology

	Mean		SD		Range
	Study 1a	Study 2b	Study 1	Study 2	Study 1	Study 2
Age	13.57	12.6	3.47	2.59	6.54 –19.04	8.00 – 16.95
Income-to-needs Ratio	3.03	3.26	2.70	2.69	.13 – 8.33	.09 – 10.35
Food Insecurity	1.89	.99	2.23	1.66	.00 – 7.00	.00 – 6
BRIEF Behavioral Regulation Index	20.00	—	13.60	—	.00 – 50.00	—
Violence Exposure	0	5.12	1.52	3.61	−.89 – 6.36	.00 – 14.00
Arrows: Baseline Time (seconds)	24.96	17.94	10.15	4.54	8.00 – 52.18	10.00 – 32.47
Arrows: Inhibitory Control (seconds)	11.35	9.12	9.74	6.41	−1.05 – 66.81	−3.88 – 39.7
Arrows: Switching Ability (seconds)	17.29	11.88	11.5	6.39	3.21 – 62.43	−22.48 – 40.35
Emotional Stroop (milliseconds)	—	60.62	—	81.72	—	−238.7 – 754
Internalizing Symptoms (T-score)	58.23	60	10.47	10.76	38 – 87	33 – 87
Externalizing Symptoms (T-score)	55.95	56.65	10.33	10.79	34 – 83	33 – 81

^a.Study 1 n=94

^b.Study 2 n=259

Measures

Socioeconomic status (SES).

A parent or guardian completed an SES questionnaire. Parent education was coded as the highest educational attainment of any parent. Income-to-needs ratio (INR) was calculated by dividing total household income by the 2014 U.S. census-defined poverty line for a family of that size, with a value less than one indicating that a family was living below the poverty line (e.g., $24,063 for a family of four) (U.S. Census Bureau, 2014). Being below the poverty line was a separate variable in order to isolate the lowest end of the income distribution. Food insecurity was used as a marker of more severe material deprivation, consistent with prior work on SES and child mental health (McLaughlin, Green, et al., 2012). Food insecurity in the past 12 months was assessed using items drawn from the short form of the U.S. Department of Agriculture’s Food Security Scale, a validated measure of food insecurity that has been used in epidemiological surveys of youth psychopathology (McLaughlin, Green, et al., 2012). See Appendix A for a description of the items and Table 1 for descriptive statistics of SES.

Psychopathology.

Internalizing and externalizing psychopathology were assessed using the Youth Self Report (YSR) form and the Child Behavioral Checklist (CBCL) (Achenbach, 1991; Rescorla & Achenbach, 2001). The internalizing and externalizing scales have demonstrated validity in discriminating between youths with and without psychiatric disorders (Achenbach, 1991; Chen et al., 1994; Kendall et al., 2007). We examined symptoms based on the higher score between the child and parent reported psychopathology in order to capture unique information contributed by both reporters (Cantwell et al., 1997). See Table 1 for descriptive statistics on psychopathology.

Parent Report of Executive Functioning.

Parent ratings of EF in Study 1 were measured using the Behavior Rating Inventory of Executive Function (BRIEF) – Parent Form (Gioia et al., 2000). This 86-item questionnaire asks parents to rate their child on everyday behavioral examples of EF. Here, we focus on BR, an EF domain that was not captured well in our behavioral task. The Behavioral Regulation Index (BRI) is a composite of the subdomains inhibition, shifting, and emotional control, and is reported as a total sum score, with a higher score indicating more difficulty with BR. As a measurement of executive functioning, the BRIEF is psychometrically sound with a high test-retest reliability (r = .81) during a two-week period, excellent convergent validity with other behavioral indications of EF (McCandless & O’ Laughlin, 2007), and strong discriminant validity when used to assess children with and without clinically significant EF difficulties (e.g., in ADHD) (Gioia et al., 2002; Reddy et al., 2011).

Executive Functioning - Arrows Inhibition Task.

Two components of executive functioning—inhibition and cognitive flexibility—were measured through the Arrows Inhibition Test (Arrows Task; Brooks, Sherman, & Strauss, 2010), which is composed of three timed trials: baseline, inhibition, and switching. See Appendix A for a description of each trial. Inhibitory control and switching ability were represented by difference in response time on each block relative to baseline. Smaller differences indicated better performance on inhibitory and switching trials. The inhibition trial of the Arrows Task has a test-retest reliability of .79 (ages 5–6), .81 (ages 7–8), .66 (ages 9–10), .80 (ages 11–12), and .82 (ages 13–16). The rule-switching trial has a test-retest reliability of .82 (ages 7–8), .78 (ages 9–10), .75 (ages 11–12), and .93 (ages 13–16) (Brooks et al., 2010). Test-retest intervals averaged 21 days (range 12–51 days). See Table 1 for descriptive statistics of the EF measures.

Violence Exposure.

Violence exposure was operationalized as exposure to child physical abuse, sexual abuse, or community violence, following prior work in this sample (Lambert et al., 2017a). Physical and sexual abuse were measured through the Child Trauma Questionnaire (CTQ), which has demonstrated high internal consistency, test-retest reliability over a two- to six-month interval, and convergent and discriminant validity with other measures (Bernstein et al., 1997, 2003). The CTQ physical abuse and sexual abuse scales demonstrated high reliability within this sample (α=.82 and α=.95, respectively) and were summed to create a child abuse score. Exposure to community violence was measured through the Screen for Adolescent Violence Exposure (SAVE), which has demonstrated high internal consistency, test-retest reliability after a two-week interval, and convergent and discriminant validity with local crime data (Hastings & Kelley, 1997). The SAVE demonstrated high reliability within this sample (α=.82). A community violence score was calculated by the sum of 12 items assessing direct exposure to community violence (e.g., being mugged or seeing someone get shot). A total violence exposure score was calculated as the sum of the standardized scores of the CTQ child abuse score and the SAVE community violence score (Lambert et al., 2017b).

Statistical Methods

Using linear regression, we first tested the association of each of the four measures of SES with internalizing and externalizing psychopathology separately (‘a’ path), for a total of eight models. Second, we tested the association of each measure of SES with the three measures of EF separately (‘b’ path) for a total of twelve models. Third, we tested each measure of EF with internalizing and externalizing psychopathology separately (‘c’ path) for a total of six models. Age and sex were included as covariates in all models. Age and sex did not moderate the associations of SES with EF or EF with psychopathology. Tests of indirect effects of SES on psychopathology through EF were conducted for each model listed above in which significant direct effects were observed for each of the three paths. Indirect effects were estimated using a non-parametric bootstrapping approach. Bootstrapped confidence intervals were calculated at 95 percent with 1,000 bootstrap samples. Bootstrapping was only performed for the mediation models. This method is recommended when testing indirect effects, which are often non-normal in distribution and thus, violate the assumptions of conventional, parametric, regression tests. The linear regression models testing the direct effects did not violate this assumption, so a standard significance test was appropriate in those cases. All modeling was performed in R v3.5.1(R Core Team, 2018) using the package ‘lavaan’ (v0.6.3; (Rosseel, 2012)). Full information maximum likelihood estimation was used to account for missing data in all models. To determine whether these associations persisted after adjusting for exposure to violence, we conducted a sensitivity analysis controlling for violence exposure in all models.

Results

SES and Psychopathology

Demographic characteristics of the sample, along with the distribution of SES, EF, and psychopathology variables are presented in Table 1, and bivariate correlations are presented in Table 2. In Study 1, lower SES was associated with higher internalizing symptoms across multiple measures of SES, including parent education, INR, and food insecurity. Lower INR and higher food insecurity, but not parent educational, were associated with higher externalizing symptoms (Table 3). After controlling for violence exposure, these associations persisted for income and parent education (Appendix Table A.1).

Table 2.

Zero-Order Correlations Among Socioeconomic Status, Violence Exposure, Executive Functioning and Internalizing and Externalizing Psychopathology Variables

Study 1
	1	2	3	4	5	6	7	8	9	10
1. Highest Parental Education	–
2. Income-to-needs Ratio	.68**	–
3. Food Insecurity	−.52**	−.62**	–
4. Below the Poverty Line	−.53**	−.71**	.64**	–
5. BRIEF Behavioral Regulation Index	−.28*	−.26*	.33**	.10	–
6. Arrows: Inhibitory Control	−.25*	−.11	.07	.06	.17	–
7. Arrows: Switching Ability	−.27*	−.23*	.23*	.14	.19	.42**	–
8. Internalizing Symptoms	−.31**	−.31**	.28**	.18	.52**	.05	.12	–
9. Externalizing Symptoms	−.18	−.34**	.28**	.21	.55**	.15	.07	.63**	–
10. Violence Exposure	−0.12	−0.12	0.15	.06	0.09	0.05	−0.14	.29**	.30**	–
Study 2
	1	2	3	4	5	6	7	8	9	10
1. Highest Parental Education	–
2. Income-to-needs Ratio	.56**	–
3. Food Insecurity	−.33**	−.50**	–
4. Below the Poverty Line	−.39**	−.64**	.51**	–
5. Arrows: Inhibitory Control	−.22*	−.24**	.26**	.35**	–
6. Arrows: Switching Ability	−.069	−.04	.11	.18*	.46**	–
7. Internalizing Symptoms	−.16*	−.22**	.34**	.17**	.16*	.09	–
8. Externalizing Symptoms	−.26**	−.29**	.37**	.22**	.31**	.25**	.67**	–
9. Emotional Stroop	.05	.16*	−.06	−.17**	−.06	−.02	.01	−.003	–
10. Violence Exposure	−.37**	−.44**	.32**	.27**	.18*	0.01	.46**	.52**	0.06	–

^*p ≤ 0.05,

^**p ≤ 0.01

Table 3.

Linear Regression Models for Socioeconomic Status and Externalizing and Internalizing Psychopathology

	Externalizing Symptoms						Internalizing Symptoms
	Study 1			Study 2			Study 1			Study 2
	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper
Income-to-needs Ratio	−.33*	−.53	−.13	−.30*	−.42	−.18	−.30*	−.50	−.10	−.23*	−.35	−.10
Highest Parent Education	−.16	−.36	.04	−.27*	−.40	−.14	−.29*	−.49	−.09	−.16*	−.30	−.03
Food Insecurity	.27*	.06	.48	.36*	.24	.47	.26*	.05	.46	.34*	.22	.45
Below the Poverty Line	.41	−.01	.84	.48*	.20	.76	.35	−.09	.78	.41*	.14	.69

Age and sex were included as covariates in all analyses;

^*95% confidence interval did not include 0.

SES and Executive Functioning

As shown in Table 4, lower SES was generally associated with difficulties with EF. Lower INR was associated with slower switching ability (i.e., higher response times on the Arrows Switching task) and lower BR (i.e., higher scores on the BRIEF BRI). Lower parent education was associated with lower inhibition (i.e. higher response time on the Arrows Task), slower switching ability, and lower BR. Higher food insecurity was associated with lower BR, but not other EF measures (Tables 4a and 4b). These findings remained after controlling for violence exposure (Appendix Tables A.2a and A.2b).

Table 4a.

Linear Regression Models for Socioeconomic Status and Executive Functioning in the Arrows Task

	Arrows: Inhibitory Control						Arrows: Switching Ability
	Study 1			Study 2			Study 1			Study 2
	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper
Income-to-needs Ratio	−0.09	−0.29	.12	−.22*	−0.36	−0.08	−0.19*	−.37	−.01	−.04	−0.20	0.11
Highest Parent Education	−0.23*	−.42	−.03	−.20*	−0.37	−0.04	−0.24*	−.43	−.05	−.06	−0.22	0.11
Food Insecurity	0.02	−.19	.23	.26*	0.12	0.41	0.16	−.03	.35	.14	−0.02	0.29
Below the Poverty Line	.10	−.32	.51	.79*	0.45	1.13	.23	−.15	0.62	.40*	0.02	0.78

Age and sex were included as covariates in all analyses;

^*95% confidence interval did not include 0.

Table 4b.

Linear Regression Models for Socioeconomic Status and Executive Functioning in the BRIEF Behavioral Regulation Index and Emotional Stroop Task

	BRIEF Behavioral Regulation Index			Emotional Stroop
	Study 1			Study 2
	β	CILower	CIUpper	β	CILower	CIUpper
Income-to-needs Ratio	−0.24*	−.45	−.03	.16*	0.03	0.28
Highest Parent Education	−0.24*	−.44	−.05	.06	−0.10	0.22
Food Insecurity	0.30*	.10	.49	−.07	−0.20	0.05
Below the Poverty Line	.21	−.23	.64	−.39*	−0.67	−0.11

Age and sex were included as covariates in all analyses;

^*95% confidence interval did not include 0.

Executive Functioning and Psychopathology

Lower BR was associated with higher externalizing and internalizing symptoms (Table 5). No associations were observed between either inhibition or switching ability and psychopathology. This association between BR and higher psychopathology remained after controlling for violence exposure (Appendix Table A.3).

Table 5.

Linear Regression Models for Executive Functioning and Externalizing and Internalizing Psychopathology

	Externalizing Symptoms						Internalizing Symptoms
	Study 1			Study 2			Study 1			Study 2
	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper
Arrows: Inhibitory Control	.13	−.08	.33	.31*	.16	.47	.01	−.20	.22	.18*	.03	.34
Arrows: Switching Ability	.03	−.20	.25	.27*	.11	.42	.07	−.15	.29	.11	−.04	.27
BRIEF Behavioral Regulation Index	.59*	.41	.77	—	—	—	.55*	.37	.74	—	—	—
Emotional Stroop	—	—	—	−.007	−.13	.12	—	—	—	.002	−.12	.13

Age and sex were included as covariates in all analyses;

^*95% confidence interval did not include 0.

Indirect Effects of SES on Psychopathology through EF

Tests of indirect effects were carried out when each of the three paths tested above was statistically significant (see Table 6 for indirect effects). Lower BR mediated the relationship between SES and internalizing and externalizing psychopathology. Lower BR accounted for 40% of the total effect of INR on internalizing symptoms (total b=−.30, 95% CI [−.51, −.09]), 61% of the total effect of food insecurity on internalizing (total b=.26, 95% CI [.07, .47]). Further, lower BR accounted for 36% of the total effect of INR on externalizing symptoms (total b=−.33, 95% CI [−.53, −.16] and 63% of the total effect of food insecurity on externalizing symptoms (total b=.27, 95% CI [.02, .49]).

Table 6.

Mediation Models and 95% Confidence Intervals for Indirect Effects of Socioeconomic Status on Psychopathology, Mediated by Executive Functioning

	Externalizing Symptoms 95% Confidence Interval						Internalizing Symptoms 95% Confidence Interval
	Study 1a			Study 2b			Study 1a			Study 2b
	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper	β	CILower	CIUpper
Income-to-needs Ratio	−.12*	−.26	−.003	−.05*	−.10	−.004	−.12*	−.28	−.01	−.02	−.06	.02
Highest Parent Education	—	—	—	−.05	−.11	.00	−.12	−.26	.00	−.03	−.08	.01
Food Insecurity	.17*	.06	.31	.05*	.003	.10	.16*	.06	.30	.02	−.02	.07
Family Below Poverty Line	—	—	—	.17*	.02	.31	—	—	—	.05	−.09	.20

^a.Study 1 Mediator: BRIEF Behavioral Regulation Index

^b.Study 2 Mediator: Arrow: Inhibitory Control

^*95% confidence interval did not include 0

Age and sex were included as covariates in all analyses.

When controlling for violence exposure in the mediation analyses, BR accounted for 41% of the total effect of INR on internalizing (total b=−.26, 95% CI [−.48, −.07]; Appendix Table A.4) and 43% of the total effect of parent education on internalizing (total b=−.25, 95% CI [−.44, −.06]). Lower BR accounted for 38% of the effect of INR on externalizing (total b=−.29, 95% CI [−.49, −.12]). BR also continued to have an indirect effect on food insecurity and internalizing symptoms (b=.14, 95% CI [.05, .26]) and externalizing symptoms (b=.15, 95% CI [.05, .29]).

Discussion

In Study 1, we observed that lower SES was associated with lower EF performance across multiple domains, and higher internalizing and externalizing symptoms. This effect largely remained after controlling for violence exposure. Lower BR was associated with higher externalizing and internalizing symptoms, and partially mediated the association of low SES with higher psychopathology.

Study 2

Method

Sample

The sample for Study 2 included 259 children and adolescents (54.1% male) aged 8–16 years (M=12.6, SD=2.59, Table 1). Participants described themselves as White (39%), Black (29.3%), Latino (11.2%), Asian (10.8%), and Other (9.7%). Youths were recruited for participation at schools, after-school and prevention programs, medical clinics, and the general community between January 2015 and June 2017. Recruitment efforts were targeted at recruiting a sample with variation in exposure to adversity. To do so, we recruited from neighborhoods with high levels of violent crime, from clinics that served a predominantly low-SES catchment area, and agencies that work with families who have been victims of violence to identify children with exposure to trauma, as well as a variety of sources in the general community (e.g., public transportation, libraries, after-school programs, coffee shops, etc.) to identify children without adversity experiences. Approximately half of children in the final sample had experienced some form of adversity and the other half were control participants, matched on age and sex to those with adversity experiences. Additional details on adversity exposure in the sample are reported elsewhere (Jenness et al., 2020; Weissman et al., 2019, 2020). The Institutional Review Board (IRB) approved all procedures. Informed consent was obtained from all participants in the study.

Measures

The measurement of SES indices matched the approach in Study 1. EF inhibition and rule-switching ability were repeated as in Study 1 by conducting the Arrows Task. Internalizing and externalizing psychopathology were also measured as in Study 1 through the YSR and CBCL questionnaires.

Executive Functioning - Emotional Stroop Task.

In addition to the Arrows Task, participants in Study 2 completed the emotional Stroop task, which assesses inhibition of responses to emotional stimuli (Etkin et al., 2010). On each trial, participants were asked to categorize a facial expression as happy or scared, while ignoring a word (i.e., “Happy” or “Fear”) overlaid on the facial expression. During congruent trials, the facial expression and word matched; during incongruent trials, the facial expression and word did not match. A larger difference between the mean reaction time in milliseconds of congruent and incongruent trials indicated worse inhibition.

Violence Exposure.

Children completed interview and self-report measures assessing child abuse and violence exposure; caregivers completed several questionnaire measures assessing children’s exposure to maltreatment and trauma. See Appendix A for further details on each measure. A composite reflecting violence exposure across multiple indicators was created based on prior work in this sample (Sumner et al., 2019). The composite was calculated by summing the total number of experiences of violence endorsed by the child and/or caregiver, including physical abuse, sexual abuse, emotional abuse, domestic violence, and exposure to other forms of interpersonal violence. Values ranged from 0–14 (M=5.11, SD=3.61). The violence exposure composite items demonstrated high reliability within this sample (α=.80).

Statistical Methods

The statistical methods used in Study 2 were identical to those used in Study 1.

Results

SES and Psychopathology

Demographic characteristics of the sample, along with the distribution of SES, EF, and psychopathology variables are presented in Table 1, and bivariate correlations in Table 2. Lower SES, including parent education, INR, food insecurity and poverty, was associated with higher internalizing and externalizing symptoms (Table 3). This association remained for food insecurity after controlling for violence exposure (Appendix Table A.1).

SES and Executive Functioning

Lower SES was generally associated with lower EF. Low SES in all four SES indicators was associated with lower inhibitory control. Poverty was the only SES measure associated with lower switching ability. Lower INR and poverty were associated with lower performance on the emotional Stroop task (Tables 4a and 4b). These associations between both food insecurity and poverty and inhibitory control remained when controlling for exposure to violence, as did the associations between both INR and poverty and the emotional Stroop task (Appendix Tables A.2a and A.2b).

Executive Functioning and Psychopathology

Lower EF was generally associated with higher externalizing and internalizing symptoms. Lower inhibitory control and lower switching ability were associated with higher externalizing symptoms. Lower inhibitory control was associated with higher internalizing symptoms. No association was observed between performance on the emotional Stroop task and psychopathology (Table 5). When controlling for exposure to violence, these results remain for the associations between lower inhibitory control and switching ability and higher externalizing symptoms (Appendix Table A.3).

Indirect Effects of SES on Psychopathology through EF

Lower inhibitory control mediated the association between lower SES and externalizing symptoms (Table 6). Lower inhibitory control accounted for 14% of the total effect of INR on externalizing (total b=−.36, 95% CI [−.50, −.22]), 13% of the total effect of food insecurity on externalizing (total b=.39, 95% CI [.25, .53]), and 20% of the total effect of being below the poverty line on externalizing symptoms (total b=.82, 95% CI [.47, 1.17]).

When controlling for violence exposure, inhibition had an indirect effect on INR and externalizing symptoms (b=−.04, 95% CI [−.08, −.001]; Appendix Table A.4) and poverty and externalizing symptoms (b=.13, 95% CI [.01, .25]). The indirect effect of inhibition on food insecurity and externalizing was not significant (b=.03, 95% CI [− .003, .07]).

Discussion

Similar to the findings in Study 1, we observed in Study 2 that lower SES was associated with lower inhibitory control and higher internalizing and externalizing symptoms. Lower inhibitory control partially accounted for the association of lower SES, specifically INR and poverty, with externalizing symptoms, above the effect of violence exposure.

Combined Discussion

In two studies, we examined the associations between different indicators of SES, domains of EF, and externalizing and internalizing psychopathology. Results from both studies were broadly similar. Lower SES was associated with lower EF performance, and lower SES and EF were associated with higher levels of psychopathology. In both studies, measures of EF (i.e., BR and inhibitory control) partially mediated the effect of SES on externalizing and internalizing psychopathology, although these indirect effects were attenuated after controlling for violence exposure. Importantly, however, BR and inhibitory control continued to have an indirect effect on the association between monetary indices of SES (i.e., INR, poverty) and psychopathology in both studies after controlling for exposure to violence.

Lower SES across all indices was associated with higher levels of both internalizing and externalizing psychopathology. This is consistent with prior research that shows psychopathology in youth is more common in children raised in low-SES households (Capistrano et al., 2016; McLaughlin, Costello, et al., 2012; McLoyd, 1998). The association between SES and psychopathology has been demonstrated across various measures of SES and many types of psychopathology (Brooks-Gunn & Duncan, 1997; Peverill et al., 2020; Slopen et al., 2010).

Lower SES was also associated with lower EF. BR (Study 1) and inhibitory control (Study 2) were associated with three indices of SES: INR, parent education, and food insecurity. Lower parent education and lower INR were also associated with lower cognitive flexibility (Study 1). Additionally, lower INR was associated with lower inhibitory control in emotional contexts (Study 2). These SES-related disparities in EF are consistent with prior research (Farah et al., 2006; Noble et al., 2007), including a recent meta-analysis (Lawson et al., 2018). We extend this prior literature, which has largely ignored the potential impact of co-occurring forms of adversity on the association between SES and EF, by demonstrating that these associations persist even after adjusting for violence exposure. Our approach is consistent with conceptual models that argue for the importance of controlling for co-occurring exposures when evaluating the developmental consequences of childhood adversity (McLaughlin & Sheridan, 2016). In particular, increasing evidence suggests that different dimensions of childhood adversity have unique developmental consequences that are obscured in studies that examine different types of adversity independently without controlling for co-occurring experiences (McLaughlin, Sheridan, & Lambert, 2014; Sheridan & McLaughlin, 2014). Here, we demonstrate that associations of SES with EF are not explained by co-occurring exposure to violence.

What might explain the association between SES and EF? Although the mechanisms linking SES with EF remain an area of ongoing research, existing evidence suggests that differences in environmental enrichment and cognitive stimulation in early childhood plays a role in the association of SES and EF. Differences in the degree of enrichment and cognitive stimulation in early home and school environments have been shown to vary in relation to SES (Bradley & Corwyn, 2002; Crosnoe et al., 2010; Hackman et al., 2015). For example, differences in linguistic experiences, such as family language complexity and syntactical structure, have consistently been shown to vary with SES (Huttenlocher et al., 2010; Rowe, 2012) and may play a role in the development of EF (Noble et al., 2005; Sheridan, Sarsour, Jutte, D’Esposito, & Boyce, 2012; Rosen et al., 2018). Cognitive stimulation (e.g., caregiver involvement in learning) and enrichment activities (e.g., variety in experiences outside the home) during childhood have been shown to partially account for the association between SES and EF, within cross-sectional (Amso et al., 2018; Sarsour et al., 2011) and longitudinal studies (Hackman et al., 2015; Rosen et al., 2019). Further, an intervention targeting cognitive stimulation in the home environment demonstrated improvement in EF over time in early childhood (Yousafzai et al., 2016). Collectively, this research indicates that cognitive stimulation in the early environment may be an underlying mechanism explaining the link between SES and EF.

All three components of EF—inhibition, cognitive flexibility, and BR—were negatively associated with externalizing and internalizing psychopathology. Lower EF has been associated with higher levels of externalizing psychopathology fairly consistently (Hobson et al., 2011; Schoemaker et al., 2013), while internalizing psychopathology has primarily been associated with lower inhibitory control (Goeleven et al., 2006; Joormann, 2006). Low EF performance, particularly lower BR and inhibition, could represent a transdiagnostic risk for psychopathology in youth (Snyder et al., 2019). Difficulty controlling one’s attention and impulses might increase behaviors and emotional reactions to external stimuli without previous thought or planning, which is common in youth with externalizing symptoms such as aggression (Giancola et al., 1996; Waller et al., 2017), hyperactivity (Berlin & Bohlin, 2002), and impulsivity (Olson et al., 2005).

Prior research has demonstrated mixed support regarding the extent to which difficulties with inhibitory control over emotional stimuli in particular are associated with internalizing psychopathology (Capistrano et al., 2016; Jarcho et al., 2013; Joormann & Gotlib, 2010). Our studies found support for this hypothesis using measures of inhibitory control in relation to neutral stimuli (i.e., arrows). However, inhibitory control in an emotional context was unrelated to psychopathology in Study 2, which is consistent with previous research using this task in youth (Jarcho et al., 2013). In contrast, parent-reported BR was associated with both internalizing and externalizing problems in Study 1. Parent-reported EF skills might be a particularly ecologically valid measure of inhibition related to emotional situations in daily life. Low BR and inhibitory control may potentially contribute to internalizing symptoms through difficulty inhibiting one’s attention and emotional response to irrelevant or negative stimuli (Capistrano et al., 2016). Future research might focus on determining which mechanisms uniquely link inhibitory control to internalizing psychopathology.

Our results suggest that certain domains of EF partially mediate the effect of low SES on psychopathology in youth, an effect that has been shown in other forms of early childhood adversity involving deprivation (e.g., institutionalization) (Tibu et al., 2016). The findings from Study 1 suggest that difficulty with BR may make it challenging for children to employ EF skills in order to adapt to their environment, which may lead to greater externalizing or internalizing symptoms. In Study 2, inhibitory control emerged as the most consistent mediator, partially accounting for the effect of three SES indices on externalizing symptoms. These findings suggest that low SES is associated with inhibitory control in ways that might increase externalizing, but not necessarily internalizing, symptoms.

Notably, BR and inhibitory control emerged as significant mediators of the association of food insecurity with psychopathology, suggesting that EF may play a role in linking severe forms of material deprivation to psychopathology, beyond the effects of violence exposure. Food insecurity is a relatively severe form of material deprivation and is uniquely associated with psychopathology in youth (McLaughlin, Green, et al., 2012; Slopen et al., 2010). Children’s cognitive development may be particularly impacted by inconsistent access to food. Although food insecurity is common in the U.S. (McLaughlin, Green, et al., 2012), it remains an under-studied form of adversity. Some indirect effects were attenuated after controlling for violence exposure, which may indicate that other mechanisms associated with violence (i.e., emotional reactivity, emotion dysregulation, social information processing) might play more of a role than EF (see McLaughlin & Lambert, 2017 for a review). Greater research is needed to disentangle the relative importance of these pathways across different forms of adversity in order to identify the most promising targets for early interventions.

Several limitations should be considered when interpreting these findings. Each of our two studies used a cross-sectional design and so cannot establish temporal relationships between the study variables. Without longitudinal data, we cannot disentangle the bidirectional relationship between EF and psychopathology to know whether differences in SES necessarily preceded differences in behavioral regulation or inhibitory control, which then preceded differences in externalizing and internalizing psychopathology. Indeed, mental disorders have been associated with compromised EF capacity, such as difficulty concentrating in depression and anxiety (Hallion et al., 2018; Keller et al., 2019), which could influence performance on EF tasks. Further, symptoms of psychopathology can impact youth’s behavior in ways that influence their parent’s perception of their EF skills, potentially leading to variability in parent-reported EF ability. However, the temporal order posited herein is supported by previous longitudinal research that has found prospective associations between SES and EF (Hackman et al., 2015; Rosen et al., 2019) and EF and psychopathology (Lonigan et al., 2017; Olson et al., 2005). Taken together, our indirect effects should be interpreted with caution given the cross-sectional design. Longitudinal studies examining EF as a mediator in these associations are clearly needed. Second, EF is a multidimensional construct, and we did not measure all components of EF in this study (e.g., working memory). Inhibitory control was tested for associations with SES and psychopathology in both studies, but only tested as a mediator in Study 2, given that in Study 1, inhibitory control was not associated with psychopathology. The reasons for differences in the results across the samples are unknown, but most likely reflect subtle differences in sample composition across the two studies. Notably, effect sizes for the association of SES with EF varied considerably across samples in a recent meta-analysis (Lawson et al., 2018). Third, we cannot infer severity of clinical impairment or diagnoses from these data, as we utilized the internalizing and externalizing subscales as continuous variables, as opposed to categories of diagnoses or clinically significant thresholds. Lastly, it is worth noting the time since data collection took place; Study 1 was conducted six years ago, and Study 2 four years ago. Nevertheless, the behavioral tasks and self-report surveys used to measure core study constructs remain the gold standards in the field. Thus, the time that has elapsed since data collection was completed is unlikely to meaningfully impact the inferences.

Across two socioeconomically diverse studies, we demonstrate multiple links between childhood SES, EF performance, and psychopathology in youth. Notably, EF partially mediated the associations of food insecurity, INR, and both internalizing and externalizing psychopathology, along with poverty, INR, and externalizing psychopathology, beyond the effect of co-occurring violence exposure. Lower EF performance most consistently accounted for the association between relatively severe forms of material deprivation and youth psychopathology. Understanding how growing up in a low-SES household influences cognitive and emotional development in ways that may confer risk for psychopathology is critical for identifying targets for early intervention.