Access to this full-text is provided by Wiley.
Content available from JCPP Advances
This content is subject to copyright. Terms and conditions apply.
Received: 2 August 2024
-
Accepted: 2 October 2024
DOI: 10.1002/jcv2.12297
ORIGINAL ARTICLE
Beyond the income‐achievement gap: The role of individual,
family, and environmental factors in cognitive resilience
among low‐income youth
Divyangana Rakesh
1,2
|Ekaterina Sadikova
3
|Katie McLaughlin
2,4
1
Neuroimaging Department, Institute of
Psychiatry, Psychology & Neuroscience, King's
College London, London, UK
2
Department of Psychology, Harvard
University, Cambridge, Massachusetts, USA
3
Department of Social and Behavioral
Sciences, Harvard T.H. Chan School of Public
Health, Boston, Massachusetts, USA
4
Ballmer Institute, University of Oregon,
Eugene, Oregon, USA
Correspondence
Divyangana Rakesh, 10 Cutcombe Road,
London, SE5, UK.
Email: divyangana.rakesh@kcl.ac.uk
Funding information
National Institute of Mental Health, Grant/
Award Numbers: R01‐MH106482, R37‐
MH119194; Action for Boston Community
Development; National Institutes of Health,
Grant/Award Numbers: U01DA041048,
U01DA050989, U01DA051016,
U01DA041022, U01DA051018,
U01DA051037, U01DA050987,
U01DA041174, U01DA041106,
U01DA041117, U01DA041028,
U01DA041134, U01DA050988,
U01DA051039, U01DA041156,
U01DA041025, U01DA041120,
U01DA051038, U01DA041148,
U01DA041093, U01DA041089,
U24DA041123, U24DA041147
Abstract
Background: Low socioeconomic status is associated with lower cognitive perfor-
mance and long‐term disparities in achievement and success. However, not all
children from low‐income backgrounds exhibit lower cognitive performance.
Characterizing the factors that promote such resilience in youth from low‐income
households is of crucial importance.
Methods: We used baseline data from participants in the lowest tertile of income‐
to‐needs in the Adolescent Brain Cognitive Development study and machine
learning to identify the factors that predict fluid and crystallized cognitive resilience
among youth from low‐income backgrounds. Predictors included 164 variables
across child characteristics, family and developmental history, and environment.
Results: Our models were reliably able to predict resilience but were substantially
more accurate for crystallized cognition (AUC =0.75) than for fluid cognition
(AUC =0.67). Key predictors included developmental factors such as birthweight
and duration of breastfeeding, neighborhood‐level factors (e.g., living in concen-
trated privilege, enrollment in advanced placement courses), children's own
temperament and mental health, and other factors such as physical activity and
involvement in extracurricular activities.
Conclusion: Our findings highlight the importance of a multifaceted approach to
promoting cognitive resilience among children from low‐income households in
future intervention work.
KEYWORDS
ABCD study, childhood and adolescence, cognitive function, poverty, resilience, socioeconomic
status
INTRODUCTION
Low socioeconomic status (SES) is associated with lower cognitive
performance (Lawson et al., 2018; Noble et al., 2007) creating long‐
term disparities in achievement and success (Best et al., 2011).
Despite considerable efforts, the SES‐achievement gap has persisted
globally (Chmielewski, 2019). Beyond individual wellbeing and
success, childhood poverty costs the USA government >$1 trillion a
year in part due to loss of economic productivity (McLaughlin &
Rank, 2018). Importantly, not all children from low‐income back-
grounds exhibit lower cognitive performance. However, most studies
have adopted a deficit‐based approach, which overlooks this het-
erogeneity present within youth living in poverty (DeJoseph
et al., 2024). Adopting an adaptation or strength‐based framework
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, pro-
vided the original work is properly cited.
© 2024 The Author(s). JCPP Advances published by John Wiley & Sons Ltd on behalf of Association for Child and Adolescent Mental Health.
JCPP Advances. 2024;e12297. wileyonlinelibrary.com/journal/jcv2
-
1 of 14
https://doi.org/10.1002/jcv2.12297
and characterizing the factors that promote positive outcomes in
youth from low‐income households is essential to inform targeted
interventions aimed at alleviating the negative sequalae of poverty.
Although growing up in poverty carries many risks, some children
defy the odds and demonstrate positive outcomes. These children are
regarded as resilient (Sattler & Gershoff, 2019). Although resilience is
a widely studied concept in developmental science, particularly in the
context of mental health, consensus on how to define the construct is
still lacking (Masten, 2014). Widely used definitions describe resil-
ience as “doing better than others facing similar risks” (Rutter, 2006)
or doing as well peers not facing the same risks (Luthar et al., 2000),
both of which require the setting of a threshold to classify individuals
as resilient versus not, a common approach adopted in develop-
mental studies (Luthar et al., 2000; Luthar & Cicchetti, 2000; and see
studies included in Zhang et al., 2023). Importantly however, only
“high threshold resilience” (i.e., performing as well as peers not facing
the same risks) has been shown to be associated with improved long‐
term outcomes (Sattler & Gershoff, 2019). Another definition con-
siders resilience as the adaptive “capacity of a dynamic system” and
focuses on the interplay between risk and protective factors that may
promote positive outcomes in the context of adversity (Mas-
ten, 2014). Our study combines these definitions to characterize the
factors that promote high cognitive function among children from
low‐income backgrounds.
While numerous studies have investigated positive outcomes
despite adversity in the context of mental health and academic
achievement, studies on cognitive performance are less common
among adolescents. While cognitive performance is related to ac-
ademic achievement, they are only moderately correlated (Tikho-
mirova et al., 2020), indicating that they are distinct constructs.
Indeed, executive function—including working memory, flexibility,
and attention—has consistently been shown to mediate associa-
tions between SES and academic achievement (Rakesh et al., 2024)
suggesting that it lies earlier in the causal chain linking low SES
with lower achievement. Furthermore, academic achievement re-
flects socio‐emotional skills, and external factors like educational
support from family and teachers (Gruijters et al., 2024). While
several studies have focused on cognitive resilience in older adults
and clinical populations (e.g., Graham et al., 2021; Willroth
et al., 2023), research on cognitive resilience among children,
particularly those from low‐income backgrounds is limited. There-
fore, studies that examine factors that may promote higher
cognitive function in children from low‐income backgrounds are
warranted.
Numerous studies have examined individual‐level factors that
contribute to positive cognitive and academic outcomes in youth.
However, cognitive development and academic achievement are
influenced by a range of factors operating at not only the individual
but also the household, community, and school levels. For instance, at
the individual level, children's pubertal development has been linked
to executive function (Stumper et al., 2020), and temperamental
reactivity has been shown to reduce the strength of the association
between low SES and lower executive function (Raver et al., 2013).
Children's behavior and activities, such as sleep duration, screen
time, and participation in extracurricular activities are also associated
with cognitive function (Kirlic et al., 2021).
Developmental and family history also play a role in children's
cognitive outcomes. For example, birthweight, even in the normal
range (Shenkin et al., 2004) and longer duration of breastfeeding
(Horta et al., 2015) correlate positively with cognitive outcomes.
Parental psychopathology can influence parent‐child dynamics, ulti-
mately contributing to individual differences in cognitive function
(Valcan et al., 2018). Indeed, close parent‐child relationships in low‐
SES students is associated with academic resilience (Kong, 2020).
Finally, numerous aspects of the social and physical environment can
influence children's cognitive outcomes. For instance, adverse child-
hood experiences, exposure to trauma, access to greenspaces,
neighborhood walkability, traffic, and aspects of the school environ-
ment have all been shown to be associated with cognitive develop-
ment (Lund et al., 2020; Piccolo et al., 2019; Sylvers et al., 2022;
Vella‐Brodrick & Gilowska, 2022). However, few studies have
examined whether these factors moderate associations between SES
and cognitive function (Rakesh et al., 2024) and most work on these
associations examines risk and protective factors at a single level of
influence without considering others.
Importantly however, based on the definition of resilience by
Masten (2014), and recent developments in the field (Masten
et al., 2021), there is a need to understand the interplay between risk
and protective factors at multiple levels in promoting resilience.
However, most studies, including those in large samples (Yan &
Gai, 2022), have examined only a few individual factors, typically
separately. Consequently, the unique contribution of these different
individual, home, and community‐level variables remains unknown. In
addition, risk and protective factors may interact with one another.
Key points
What's known?
�Not all children from low‐income backgrounds exhibit
lower cognitive performance.
�The factors that promote such resilience in youth from
low‐income households remain unknown.
What's new?
�We used machine learning to identify the factors
(including child characteristics, family and developmental
history, and environment) that predict fluid and crystal-
lized cognitive resilience among youth from low‐income
backgrounds.
What's relevant?
�We found developmental factors such as birthweight and
duration of breastfeeding, neighborhood‐level factors
(e.g., living in concentrated privilege, enrollment in
advanced placement courses), children's own tempera-
ment and mental health, and other factors such as
physical activity and involvement in extracurricular ac-
tivities to predict cognitive resilience.
�Our findings highlight the importance of a multifaceted
approach to promoting cognitive resilience among chil-
dren from low‐income households in future intervention
work.
2 of 14
-
RAKESH
ET AL.
For example, given their roles in positive child outcomes (Goetschius
et al., 2023; O’Malley et al., 2015; Raniti et al., 2022; Vella‐Brodrick &
Gilowska, 2022), it is possible that positive school environments or
access to greenspaces may buffer the risk conferred by household‐
level factors. However, since studies examine individual factors,
and typically in separate models, such interactions remain unin-
vestigated. Further, the protective role of other factors that may
promote positive cognitive outcomes among children from low‐
income families (e.g., neighborhood walkability) have yet to be
investigated. As such, given that children live in complex socio‐
ecological environments (Bronfenbrenner, 1994), there is a need
for work that explores the independent and joint effects of a range of
different individual, home, and community and school level factors.
Such exploratory research is essential for understanding how chil-
dren's behavior, temperament, family history, and environment
contribute to cognitive resilience and for identifying modifiable tar-
gets for future interventions. Given the large number of children
affected by economic hardship (ALICE in Focus, 2022; Annie E. Casey
Foundation, 2014; Crouch et al., 2019), such investigations have
critical public health relevance.
The aim of this exploratory study was to identify factors that
promote cognitive resilience among youth from low‐income back-
grounds in a large population‐based sample. To this end, we lever-
aged the Adolescent Brain Cognitive Development (ABCD) Study, a
large population‐based cohort of children aged 9–10 years, and
machine learning models to comprehensively investigate the role of
child characteristics (including behaviors and activities, personality
and temperament, physical and mental health), family history (such as
parent mental health and developmental history), and environment
(including the home environment, neighborhood environment, and
traumatic experiences) in cognitive resilience among low‐income
children.
METHODS AND MATERIALS
Participants
Study participants were from the ABCD study (baseline assessment;
release 5.0). The ABCD study has enrolled over 11,500 children aged
9–10 years to comprehensively examine psychological and neurobio-
logical development in a large sample (Garavan et al., 2018). The
study's 21 sites were strategically chosen for demographic similarity to
the overall US population. Within each site, participants were
randomly selected from public, public charter, and private schools
within a 50‐mile radius. Written informed consent was obtained from
all caregivers, and all children provided assent. The rights of partici-
pants were protected under the local institutional review boards. The
final sample included 3373 youth that met our inclusion criteria (see
subsequent sections). These participants were 28.5% non‐Hispanic
White, 28.6% non‐Hispanic Black, 30.3% Hispanic, 0.9% Asian, and
11.6% other race/ethnicity (which includes multiracial identities).
Income‐to‐needs and cognitive performance
Income‐to‐needs. The income‐to‐needs ratio was calculated by
dividing the median value of the income band by the federal poverty
line corresponding to the household size. A value of 1 would indicate
being exactly at the poverty threshold and values above and below 1
would indicate being above and below the threshold, respectively.
Given our focus on cognitive resilience in the context of low SES, only
participants in the bottom tertile (33%) of the income‐to‐needs dis-
tribution were included in analyses.
Cognitive performance. Cognitive performance was assessed using
the NIH Toolbox Cognition Battery (Luciana et al., 2018). Age‐
corrected composite scores of crystallized and fluid cognition were
used in our analyses (Heaton et al., 2014; Luciana et al., 2018). Fluid
and crystallized cognition were analyzed in separate models as they
are distinct constructs (Cattell, 1963,1987). Fluid cognition pertains
to the capacities essential for abstract reasoning, while crystallized
intelligence encompasses culturally acquired knowledge that is more
likely to be environmentally determined (Cattell, 1963,1987). The
NIH Toolbox Cognition Battery assesses several cognitive abilities
including attention, executive function, working memory, episodic
memory, language, and processing speed (Akshoomoff et al., 2013;
Luciana et al., 2018). The tasks include the List Sorting Working
Memory, Dimensional Change Card Sort, Flanker Inhibitory Control
and Attention, Picture Sequence Memory, and Pattern Comparison
Processing Speed for assessing fluid cognitive Functioning, and Pic-
ture Vocabulary and Oral Reading Recognition for assessing crys-
tallized cognitive functioning. Importantly, performance on these
tasks has shown to be predictive of academic performance (Distefano
et al., 2023).
Cognitive resilience
Cognitive resilience is not a well‐defined concept in the literature
(Haft et al., 2016; Sattler & Gershoff, 2019). Higher cognitive
function—less common among youth from low‐income households—is
a strong predictor of academic achievement and long‐term success.
Based on Luthar et al. (2000) and recent findings by Sattler and
Gershoff (2019) showing that "high threshold resilience" is most
predictive of positive long‐term outcomes, we define resilience as the
presence of high cognitive function in children from low‐income
households. To this end, income‐to‐needs was divided into tertiles
(high, medium, low), and youth from the bottom tertile with above
average performance on fluid cognition or crystallized cognition were
categorized as resilient (N=1158 for fluid and 888 for crystallized) and
below average fluid cognition or crystallized cognition scores were
categorized as below‐average‐scoring (BAS; N=2215 for fluid and
2485 for crystallized). Individuals in the resilient group had similar
crystallized cognitive (M
Resilient
=119.32, M
High income
=113.09,
M
BAS
=89.84) and fluid cognitive (M
Resilient
=108.58,
M
High income
=100.38, M
BAS
=80.77) scores to those in the high income
tertile. See the Supplement for distributions of cognitive scores for the
three groups (eFigure 2a and 2b).
A rigid poverty threshold was not employed to identify “poor”
families as it substantially underestimates the number of children
living in poverty (e.g., 16% according to the federal poverty line).
Many households, despite being employed, cannot afford essential
costs like housing, childcare, food, healthcare, and transportation
(ALICE in Focus, 2022). Given this, and the fact that the relationship
between SES and cognitive outcomes is not confined to a specific
threshold but rather exists along a continuum (with stronger
PREDICTORS OF COGNITIVE RESILIENCE IN LOW‐INCOME YOUTH
-
3 of 14
associations at the lower end of the income distribution)
(McLoyd, 1998), we chose to use the lowest tertile of income‐to‐
needs in our analyses. Participants in the lowest tertile were
indeed from poor or near poor families, with a mean income‐to‐needs
ratio of 1.01 (SD =0.60, min =0.03, max =2.02) and most families
(95.7%) were below 200% of the poverty line. Distributions of
income‐to‐needs for those in the lowest tertile versus highest tertile
have been provided in the Supplement (eFigure 1a). Nonetheless, to
ensure that results did not depend on the chosen threshold, we
considered a quartile income threshold in sensitivity analyses.
Predictor variables
A total of 164 predictor variables assessed child characteristics,
family and developmental history, and environment (Figure 1; see
Table S1 for a full list of predictors and their corresponding assess-
ment methods). Child characteristics (41 variables) covered behav-
iors and activities (e.g., screentime, extra‐curricular activities),
temperament (e.g., impulsive and prosocial behavior), and health (e.g.,
internalizing and externalizing symptoms, pubertal development, and
sleep quality). Family and developmental history (32 variables)
included parent demographics, maternal age at birth, birthweight,
breastfeeding duration, and caregiver mental health. Finally, 91
variables assessed the child's environment, including the home
environment (e.g., family conflict, financial adversity, caregiver
warmth), history of traumatic events, and the neighborhood
environment (e.g., school environment, childhood opportunity
indices, crime rate, residential segregation, peer behaviors). The mice
package in R (Buuren & Groothuis‐Oudshoorn, 2011) was used to
impute missing values for the predictor variables through chained
equations with predictive mean matching. This procedure imputes
missing data through an iterative series of prediction models using
other variables in the data set. It is frequently used in studies that
employ machine learning analysis methods and demonstrates
consistently low root mean squared error, even in longitudinal set-
tings (Shaw et al., 2023).
Prediction models
First, the data was split into 10 folds (with children from the same
family retained in the same fold). We ran nested 10‐fold cross‐
validated machine learning classification analyses to predict crystal-
lized and fluid cognitive resilience outcomes (in separate models)
using the predictor variables, which were all included in the model
simultaneously. The first analytical step involved applying LASSO
(Least Absolute Shrinkage and Selection Operator) regression.
LASSO is a regularization technique that helps to mitigate multi-
collinearity and prevent overfitting by shrinking some regression
coefficients to zero, effectively excluding less relevant predictors
from the model. This process results in a more parsimonious model
with only the most influential variables retained. However, while
LASSO aids in predictor selection and reduces redundancy, estimates
FIGURE 1 An overview of the predictor space and visual depiction of the methodological approach. The LASSO regression for each
outcome (crystallized and fluid cognitive resilience) considers all listed potential predictors and shrinks the coefficients of variables that are
irrelevant and redundant to zero. For each outcome, LASSO‐selected predictors are then included in XGBoost models, which allow for complex
interactions and non‐linearities using a tree‐based prediction approach. Relative importance of predictor variables in the XGBoost models is
assessed using SHAP values. Non‐linearities in the relationships between selected predictors and each outcome are assessed using
dependence plots and estimates of linear relationships are summarized using odds ratios from logistic regressions.
4 of 14
-
RAKESH
ET AL.
obtained from LASSO are biased as it applies regularization. To
address this, we next utilized logistic regression with the predictors
selected by LASSO (i.e., those with non‐zero coefficients). Logistic
regression is used for binary outcome variables and provides esti-
mates of the odds ratios, which reflect the strength and direction of
the relationship between each predictor and the outcome.
Lastly, Extreme Gradient Boosting (XGBoost; R package xgboost;
Chen & Guestrin, 2016) classification was used to build a prediction
model with the LASSO‐selected predictors to incorporate potential
interactions and non‐linearities in associations and ascertain relative
variable importance (Lundberg & Lee, 2017). XGBoost builds an
ensemble of decision trees sequentially, where each tree corrects the
errors of its predecessors. This approach captures complex in-
teractions and non‐linear relationships between predictors and out-
comes. Variable importance was assessed using SHapley Additive
explanation (SHAP) values. SHAP values provide an understanding of
each predictor's contribution to the model's predictions. SHAP values
ensure that each predictor's impact is assessed by considering all
possible combinations of predictors. Each SHAP value represents
how much a specific predictor changes the model's prediction
compared to the average prediction if that predictor were not
included. By aggregating these individual contributions, SHAP values
provide a measure of each variable's overall importance. This means
that the contribution of a predictor is evaluated not in isolation, but
in the context of other predictors. Specifically, positive and negative
SHAP values indicate that the feature increases and decreases the
predicted value, respectively. SHAP values inherently capture non‐
linearity and can thus be used to examine the dependence of the
prediction on each predictor across the range of its values in
dependence plots. See Figure 1for a visual depiction of the approach.
The above approach was selected from a number of strategies
including the direct application of XGBoost and SuperLearner
ensemble classification to the full set of predictors. Since accuracy
did not differ substantially between strategies (see Table S2), results
from LASSO regression followed by XGBoost classification are re-
ported in the main manuscript due to maximal interpretability and
accuracy (as assessed through area under‐the‐curve [AUC]).
Sensitivity analyses
In addition to testing a more conservative threshold to define the
low‐income group (namely the lowest quartile rather than lowest
tertile of the income‐to‐needs distribution), we also repeated the
main analysis to discriminate between those with above average
cognitive function and those with below average cognitive function
among youth from high income backgrounds (i.e., highest income
tertile). This analysis allowed us to determine which factors play a
role in promoting cognitive function specifically in a low‐income
context, as opposed to promoting high cognitive function generally.
Lastly, we ran a sensitivity analysis accounting for clustering within
families. The code for all analyses conducted is shared in the following
repository: https://github.com/katsadikova/ABCD_cog_resilience.git.
RESULTS
Demographic information
The sample used in the main analysis comprised 3373 children with a
mean age of 9.47 �0.51 years. See Table 1for demographic infor-
mation on resilient and BAS children.
Model performance
The LASSO regression model had good discrimination between
resilient and BAS children but was substantially more accurate for
crystallized cognition (AUC =0.75) than for fluid cognition
(AUC =0.67).
Predictors of cognitive resilience
Overall, 35 of the 164 considered factors across child characteristics,
family and developmental history, and environment contributed
meaningfully to the prediction of both fluid and crystallized cognitive
resilience (see Figure 2for SHAP values representing relative pre-
dictor importance and Table 2for odds ratios for LASSO‐selected
predictors).
Common predictors for fluid and crystallized cognitive
resilience
Of the 35 LASSO‐selected predictors, 14 were common between
crystallized and fluid cognitive outcomes and included six child
characteristics, four family history markers, and four environmental
TABLE 1Sample descriptives.
Fluid cognition Crystallized cognition
Resilient BAS Resilient BAS
n(%) 1158 (34.3%) 2215 (65.7%) 888 (26.3%) 2485 (73.7%)
Age, years 9.47 �0.51 9.45 �0.51 9.49 �0.51 9.45 �0.51
Female sex 48% 49% 44% 50%
Income‐to‐needs 1.12 �0.58 0.95 �0.59 1.21 �0.56 0.94 �0.59
Fluid cognitive functioning 105.09 �17.39 93.69 �14.67 119.32 �12.48 89.84 �9.38
Crystallized cognitive functioning 108.59 �10.43 80.78 �10.50 98.82 �17.05 87.29 �15.71
Total cognitive functioning 107.96 �13.42 84.38 �12.22 110.58 �14.19 86.00 �12.50
Note: Values indicate mean �standard deviation unless otherwise specified. BAS =below‐average‐scoring.
PREDICTORS OF COGNITIVE RESILIENCE IN LOW‐INCOME YOUTH
-
5 of 14
factors (Figure 3). Among child characteristics, higher frequency of
physical activity and participation in performing arts activities were
associated with greater odds of resilience. Lower levels of positive
urgency and lower attention symptoms were found to increase the
odds of resilience. Finally, older children were also more likely to be
classified as resilient.
As for family history, longer duration of breastfeeding and higher
birthweight were associated with higher odds of resilience. However,
the benefits with respect to crystallized cognition plateaued after
5 months of breastfeeding, while the potential benefits to fluid
cognition persisted to approximately 15 months. Further, while pri-
mary and secondary caregiver educational attainment were both
relevant for resilience, the former was more important for fluid
cognition and the latter for crystallized cognition.
At the neighborhood level, lower public assistance rates and
living in higher privilege also predicted resilience. Finally, higher in-
come even within the bottom tertile had benefits for resilience,
particularly with respect to crystallized cognition.
Unique predictors of fluid cognitive resilience
Resilience in the context of fluid cognition was uniquely predicted by
less advanced pubertal development, stricter parental rules around
substance use in the household, and lower vacancy and poverty rates
in the neighborhood (Figure 4).
Unique predictors of crystallized cognitive resilience
A much larger number of factors predicted resilience uniquely in the
context of crystallized cognition (Figure 5). Among child characteristics
and behavior, participation in visual arts improved the odds of resil-
ience. Other relationships were slightly more complex. Lower negative
urgency, some sensation seeking and some degree of spontaneity – or
lack of planning – promoted resilience. However, higher self‐ratings of
prosocial behaviors, reporting a large number of close friendships
(more than 20), experiencing higher levels of social problems, and the
complete absence of parent‐reported youth anxiety or depression
symptoms were associated with lower odds of resilience.
In terms of family history, children with a parent reporting higher
levels of attention symptoms and seeking assistance from a mental
health professional were associated with crystallized cognitive
resilience. Lastly, for environment, lower family conflict, living in
neighborhoods with higher enrollment in advanced placement cour-
ses, higher neighborhood employment rates, stronger third grade
reading proficiency, and lower school poverty promoted greater odds
of crystallized cognitive resilience. However, self‐reported highly
FIGURE 2 Variable importance for predicting resilience for fluid (left) and crystallized cognitive (right) across three domains. The three
colors represent the three domains of the predictors: teal for child characteristics, yellow for family and developmental history, and blue for
environment. This figure displays the absolute SHAP value for each predictor to indicate variable importance. SHAP values provide an
understanding of each predictor's contribution to the model's predictions. Each SHAP value represents how much a specific predictor changes
the model's prediction compared to the average prediction if that predictor were not included. Positive and negative SHAP values indicate that
the feature increases and decreases the predicted value, respectively. For each feature, average absolute SHAP values across individuals
reflect the marginal contribution of the feature across all possible feature combinations, thus serving as a quantitative measure of its
importance.
6 of 14
-
RAKESH
ET AL.
TABLE 2Odds ratios for predictors of fluid and crystallized cognitive resilience among children from households within the lowest INR
tertile (N=3373).
Fluid cognitive resilience
Crystallized cognitive
resilience
Child characteristics OR 95% CI POR 95% CI P
Behavior and activities Participation in performing arts
1
1.28*(1.09,1.50) <0.001 1.28*(1.07,1.53) 0.007
Participation in visual arts
2
1.32*(1.06,1.64) 0.01
Temperament Behavioral activation ‐drive 0.97*(0.95,1.00) 0.028 0.98 (0.96,1.01) 0.28
Positive urgency 0.96*(0.94,0.99) <0.001 0.97*(0.94,1.00) 0.04
Negative urgency 0.96*(0.93,1.00) 0.04
Lack of planning 1.04*(1.00,1.08) 0.03
Sensation‐seeking 1.06*(1.03,1.10) <0.001
Prosocial behavior (youth report on self)
2
0.72*(0.56,0.91) 0.006
Mental health CBCL: Anxious/depressed 1.08*(1.04,1.11) <0.001
CBCL: Social problems
2
0.92*(0.87,0.97) 0.002
CBCL: Attention problems 0.93*(0.91,0.95) <0.001 0.91*(0.88,0.94) <0.001
Number of close friends 0.99*(0.98,1.00) 0.03
Physical health and attributes Age (months) 1.02*(1.01,1.03) <0.001 1.01*(1.00,1.02) 0.03
Pubertal development score
1
0.89*(0.81,0.97) 0.006
Physical activity
1,2
1.04*(1.01,1.08) 0.01 1.06*(1.02,1.10) 0.002
Family history
Demographics Primary caregiver educational attainment 1.05*(1.00,1.1) 0.03 1.04 (0.99,1.10) 0.12
Secondary caregiver educational attainment 1.02 (0.98,1.06) 0.35 1.08*(1.03,1.13) <0.001
Developmental history Birthweight 1.11*(1.06,1.18) <0.001 1.15*(1.09,1.23) <0.001
Breastfeeding (months) 1.01 (1.00,1.02) 0.08 1.01*(1.00,1.02) 0.04
Parent attention problems
2
1.04*(1.02,1.05) <0.001
Mother seen a mental health professional 1.21 (1.00,1.47) 0.05
Environment
Home environment Income‐to‐needs
1
1.16*(1.01,1.33) 0.04 1.5*(1.28,1.76) <0.001
Family conflict (youth‐reported)
2
0.94*(0.90,0.98) 0.007
Lax parental rules on substances
1
0.91*(0.86,0.97) <0.001
Neighborhood School environment 0.96*(0.94,0.99) 0.02
Neighborhood advanced placement enrollment
1,2
1.94*(1.32,2.87) <0.001
Neighborhood school poverty
2
1.00 (0.99,1.00) 0.83
Neighborhood reading proficiency (3rd grade) 1.00 (1.00,1.00) 0.40
Neighborhood housing vacancy rate
1
0.98*(0.97,1.00) 0.02
Neighborhood industrial air pollutants
2
1.06*(1.03,1.09) <0.001
Neighborhood employment rate
2
1.01 (1.00,1.02) 0.08
Neighborhood poverty rate
1
0.99 (0.98,1.00) 0.21
Neighborhood public assistance rate
1,2
1.00 (0.99,1.01) 0.42 1.00 (0.99,1.01) 0.71
Index of Concentration at the Extremes (Income)
2
1.01 (0.53,1.93) 0.97 1.18 (0.60,2.32) 0.64
Index of Concentration at the Extremes (Income þRace)
1,2
1.14 (0.55,2.36) 0.72 1.33 (0.56,3.17) 0.52
Abbreviation: CBCL, Child Behavior Checklist.
*
Two‐sided p‐value <0.05.
Based on results from the sensitivity analysis:
1
Predictors for fluid cognitive resilience in the low‐income, but not in the high‐income analysis.
2
Predictors for crystallized cognitive resilience in the low‐income, but not in the high‐income analysis.
PREDICTORS OF COGNITIVE RESILIENCE IN LOW‐INCOME YOUTH
-
7 of 14
FIGURE 3 Dependence plots for predictors common to fluid (left) and crystallized cognitive resilience (right). A dependence plot visualizes
the relationship between a single predictor's value (shown on the x‐axis of each panel) and its SHAP value (shown on the y‐axis), which
represents the impact of that predictor on the model's prediction. In these plots, a higher SHAP value indicates that the predictor is positively
contributing to cognitive resilience, while a lower SHAP value suggests a negative contribution. For example, if a plot shows that as the
predictor value increases, the SHAP value also increases, this suggests that higher values of that predictor are associated with greater
cognitive resilience. The three colors represent the three domains of the predictors: teal for child characteristics, yellow for family and
developmental history, and blue for environment.
FIGURE 4 Dependence plots for predictors unique to fluid cognition resilience. A dependence plot visualizes the relationship between a
single predictor's value (shown on the x‐axis of each panel) and its SHAP value (shown on the y‐axis), which represents the impact of that
predictor on the model's prediction. In these plots, a higher SHAP value indicates that the predictor is positively contributing to cognitive
resilience, while a lower SHAP value suggests a negative contribution. For example, if a plot shows that as the predictor value increases, the
SHAP value also increases, this suggests that higher values of that predictor are associated with greater cognition resilience. The colors
represent the domains of the predictors: teal for child characteristics and blue for environment.
8 of 14
-
RAKESH
ET AL.
positive school environments were associated with lower odds of
resilience.
Sensitivity analyses
We found the results to be stable when the low‐income group was
defined as those in the bottom quartile (rather than tertile) of the
income‐to‐needs distribution. Of the top 10 predictors identified in
the main analysis, 9 were also selected as important predictors in the
sensitivity analysis for both fluid and crystallized cognition (see the
Tables S5a and S5b in the Supplement for the list of predictors).
Comparisons of predictors of above‐average crystallized and
fluid cognition in the low‐income and high‐income samples are
summarized in the Supplement in Tables S4a and S4b, respectively.
Our findings revealed that certain factors, such as caregiver educa-
tion levels and the duration of breastfeeding, were associated with
high cognitive performance in both income groups. However, several
factors were unique to the low‐income group (marked with footnotes
in Table 2). Among the factors that were common to both low‐income
and high‐income analyses, they were less prevalent or, in the case of
negative factors, more prevalent in the low‐income group (e.g.,
duration of breastfeeding is longer in high‐income families), evi-
denced by the mean and SD values reported in Table S4a and S4b.
Correcting the main analysis for clustering within families
demonstrated little impact on the estimated odds ratios and their
standard errors (compare results reported in Table 2to those re-
ported in Table S3 in the Supplement).
DISCUSSION
The goal of this study was to comprehensively characterize pre-
dictors of cognitive resilience during late childhood. Using aspects of
youth behavior and activities, family and developmental history, and
the environment, we were able to successfully discriminate between
FIGURE 5 Dependence plots for predictors unique to crystallized cognitive resilience. A dependence plot visualizes the relationship between
a single predictor's value (shown on the x‐axis of each panel) and its SHAP value (shown on the y‐axis), which represents the impact of that
predictor on the model's prediction. In these plots, a higher SHAP value indicates that the predictor is positively contributing to cognitive
resilience, while a lower SHAP value suggests a negative contribution. For example, if a plot shows that as the predictor value increases, the SHAP
value also increases, this suggests that higher values of that predictor are associated with greater cognitive resilience. The three colors represent
the three domains of the predictors: teal for child characteristics, yellow for family and developmental history, and blue for environment.
PREDICTORS OF COGNITIVE RESILIENCE IN LOW‐INCOME YOUTH
-
9 of 14
cognitively resilient and BAS children. The predictive accuracy and
discrimination achieved by our primary modeling strategy is compa-
rable to other work employing machine learning strategies for clas-
sification problems in social and behavioral science (Papini
et al., 2023). Prediction accuracy was higher for crystallized than fluid
cognitive resilience and a larger number of predictors were found to
predict crystallized cognitive resilience.
To identify factors that specifically enhance cognitive function in
low‐income contexts, rather than those that generally support high
cognitive performance, we conducted additional analyses focusing on
high‐income youth (i.e., the highest income tertile) to differentiate
between individuals with above‐average and below‐average cogni-
tive function. These analyses revealed that while some factors, such
as primary and secondary caregiver education and longer breast-
feeding duration, contribute to high cognitive performance across
both income groups, several factors were uniquely predictive of
cognitive resilience in the low‐income group. These factors included
participation in visual and performing arts, pubertal development,
physical activity, income‐to‐needs ratio, family conflict, and various
neighborhood‐level factors, such as poverty rate, school poverty, and
advanced placement enrollment. These factors could be considered
as resilience‐promoting factors, as they seem to specifically enhance
cognitive function in low‐income settings. However, even among the
factors that were common across income levels, we observed dis-
parities in their prevalence between low‐and high‐income groups.
For instance, positive factors like longer breastfeeding duration were
less common among low‐income youth, while negative factors, such
as mental health problems, were more prevalent. This underscores
the importance of promoting these factors in low‐income contexts,
even though they are beneficial across income levels, and are
therefore discussed in this paper.
Several factors predicted both fluid and crystallized cognitive
resilience. Notably, even though all children were in the lowest tertile
of income‐to‐needs, small incremental increases in income‐to‐needs
had a discernible association with resilience. This is consistent with
studies showing that changes in income have particularly meaningful
effects on children's cognitive and schooling outcomes at the lower
end of the income spectrum (Cooper & Stewart, 2013). Similarly,
consistent with prior work (Davis‐Kean, 2005), higher caregiver
educational attainment was associated with higher odds of resilience.
Other factors included children's temperament, attention problems,
extracurricular activities, and developmental history—including
breastfeeding, and birthweight. Importantly, more frequent moder-
ate to intense physical activity and participation in performing arts
increased the odds of cognitive resilience. Our findings are consistent
with intervention studies showing the benefits of physical activity on
cognitive outcomes (Haverkamp et al., 2020) and highlight the
importance of physical activity, particularly among children from low‐
income backgrounds (Foster & Marcus Jenkins, 2017). Even after
accounting for a wide range of other factors, participation in visual
and performing arts was associated with resilience, which stands in
contrast to prior null findings (Foster & Marcus Jenkins, 2017).
Future intervention studies should investigate whether participation
in extracurricular activities supports cognitive function among chil-
dren from low‐income families.
Children with higher attention problems and lower birthweight
were less likely to be classified as resilient. Given that attention
problems are more prevalent in low‐SES youth (Russell et al., 2016),
there is a need for targeted interventions specifically designed for
children grappling with attention issues. Moreover, future work
should aim to pinpoint effective strategies to promote positive
cognitive outcomes among children born with low birthweight.
Finally, longer duration of breastfeeding was also associated with
higher odds of cognitive resilience. In line with advice by the World
Health Organization, our findings highlight the importance of
providing support to women from low‐income households post birth
(e.g., paid maternity leave, dedicated spaces for breastfeeding in the
workplace and public places, flexible working days) and increasing
public messaging around the importance of breastfeeding to improve
cognitive outcomes among children from low‐income households.
Fluid cognitive resilience was uniquely predicted by a few factors
including pubertal development, parental rules on substance use, and
neighborhood‐level factors. Specifically, less advanced pubertal sta-
tus was associated with higher odds of fluid cognitive resilience. This
finding aligns with the hypothesis that exposure to chronic stress,
which is more common in low‐income contexts, can trigger faster
pubertal development (Ellis et al., 2022), contributing to lower
cognitive performance (Stumper et al., 2020). Further, stricter
parental rules on substance use are positively associated with fluid
cognitive resilience. Rather than being a mechanism of resilience,
higher parental monitoring regarding substance use may be a proxy
for other factors (e.g., parental expectations) that support cognitive
resilience (Yan & Gai, 2022) but were not measured in the present
study.
Factors across all three domains uniquely predicted crystallized
cognitive resilience. For example, lower levels of social problems and
having a small circle of close friends was associated with higher odds
of resilience. Previous research has shown that peer support pro-
motes resilient functioning in psychosocial domains (Harmelen
et al., 2017), which may in turn contribute to the positive cognitive
outcomes observed in this study. This aligns with work showing links
between school connectedness and belonging and wellbeing (Raniti
et al., 2022). Future work should directly test whether promoting
greater integration in classrooms and facilitating friendships through
play and activities during and after school benefits the cognitive
outcomes of children from low‐income households. Additionally,
lower levels of family conflict were associated with greater odds of
crystallized cognitive resilience. Consistent with prior work that in-
dicates that stress could influence the development of physiological
systems that regulate attention (Hinnant et al., 2013), our results
suggest that a family environment characterized by conflict may
hinder a child's ability to focus, learn, and retain information.
Various neighborhood‐level factors emerged as predictors of
fluid and crystallized cognitive resilience. Specifically, living in
neighborhoods characterized by higher privilege was associated with
higher odds of resilience across both domains. Further, lower levels
of neighborhood and school poverty, housing vacancy rates, and
neighborhood‐level academic proficiency—including advanced
placement enrollment and third‐grade proficiency—were associated
with higher odds of crystallized cognitive resilience. These findings
underscore the importance of neighborhood environments for
cognitive development (Lloyd et al., 2010), which likely influence
child development through multiple pathways. For example, in
addition to structural community characteristics such as quality of
10 of 14
-
RAKESH
ET AL.
education, advantaged neighborhoods are also associated with
parenting behaviors such as the utilization of education‐focused
practices (Burton & Robin, 2000; Greenman et al., 2011; Klebanov
et al., 1994; Shumow & Lomax, 2009), which may contribute to
cognitive resilience among low‐income children. Further, the pres-
ence of high‐achieving peers who excel in reading and enroll in
advanced placement classes can model behaviors that foster crys-
tallized cognitive resilience. Future efforts could explore how
enhancing access to early childhood education programs and
advanced placement courses, particularly in resource‐deprived
neighborhoods, might support the development cognitive resilience
among children from low‐income backgrounds.
Some findings were more challenging to interpret. For example,
fluid cognitive resilience was associated with not having very high
levels of behavioral drive, positive urgency, and planning. Further, not
having very low levels of sensation seeking was also associated
higher odds of resilience. These findings demonstrate that children
from low‐income households who are open to some degree of nov-
elty and stimulation, less impulsive in response to positive emotions,
and exhibit moderate or balanced level of motivation, may demon-
strate better cognitive resilience. Further, the dependence plots
indicate that these associations are likely driven by non‐linearities at
the tails of the predictor distributions but are close to null when the
values of these predictors that are most common in the sample are
considered. Similarly, our results show that having a primary care-
giver with higher self‐reported attention symptoms and a history of
mental health professional consultations increases odds for crystal-
lized cognitive resilience. These factors likely do not function as
direct mechanisms of resilience. Instead, it is possible that children in
households where the primary caregiver acknowledges and ad-
dresses their own mental health challenges—manifested through
attention problems—and is proactive in seeking professional support
tend to fare better cognitively.
Some limitations should be considered when interpreting findings.
First, the study was cross‐sectional. Therefore, we cannot comment
on directionality or causality and any clinical implications are purely
speculative. Longitudinal and intervention studies with rigorous
adjustment for measured confounders and assessment of the influ-
ence of unmeasured confounding within a causal framework are
needed. Of note, the present study leveraged baseline data as it was
not possible to use composite scores from post‐baseline visits since
the entire cognitive battery was not administered at subsequent time‐
points (see ABCD release notes). Second, it is unknown whether
resilience is transient or persistent. Future longitudinal work should
assess factors that predict cognitive resilience across time‐points.
Third, given the major role of neighborhood level factors, it is un-
known whether these findings generalize. Similar investigations will
need to be conducted using data from other populations. Four, the
odds ratios from the logistic regression models need to be interpreted
with caution in instances where non‐linearities are evident in
dependence plots. Such odds ratios are nonetheless useful for
approximating effect sizes. Five, some of the identified associations
were challenging to interpret – for example, evidence that odds of
resilience are higher in the most supportive school environments or in
neighborhoods with higher airborne pollutant levels. These findings
may be due to measurement error or skew in predictor distributions
(e.g., most children living in urban polluted environments). Future
research should aim to characterize these associations more thor-
oughly. Six, several studies, including ours, have examined associa-
tions between SES and brain structure and function (Rakesh
et al., 2022,2023; Rakesh & Whittle, 2021; Rakesh, Zalesky, &
Whittle, 2021). Evidence suggests that positive environmental factors
can mitigate the impact of low SES on brain outcomes (Brody
et al., 2019; Rakesh, Seguin, et al., 2021; Whittle et al., 2017), and
distinct neurobiological profiles have also been identified in youth
from low‐income backgrounds who show positive outcomes (Ellwood‐
Lowe et al., 2022). However, few studies have investigated the
neurobiology underlying cognitive resilience among low‐income
youth. Future work should use a machine learning approach to iden-
tify brain features that may predict such resilience. Given that so-
cioeconomic gaps in achievement have not declined over several
decades despite substantial efforts, this research is crucial for iden-
tifying novel targets for intervention. Seven, while efforts were made
to make the ABCD sample representative, parents in this sample have
higher income‐to‐needs and educational attainment. As such, the
bottom tertile of the ABCD sample may not reflect the bottom tertile
of the national population, thus impacting the generalizability of the
results. Eight, although the NIH toolbox includes seven tasks that
measure a range of cognitive abilities, including attention, executive
function/cognitive flexibility, working memory, episodic memory,
language, and processing speed, and are associated with academic
performance (Distefano et al., 2023), factors that promote resilience
in other cognitive domains (e.g., sustained attention) may have been
missed. Nine, our study focused on a specific set of predictors. There
are several other factors that may promote cognitive resilience such
as access to mental health services, nutrition, social support networks,
and special needs education that were not assessed in the present
sample. Future work should consider these and other factors we did
not assess as predictors of cognitive resilience. Ten, the machine
learning methods used in our study, while powerful in identifying
patterns and relationships within the data, are inherently atheoretical.
Future research could benefit from an approach that integrates ma-
chine learning with theory‐driven hypotheses. This would involve
using machine learning to identify potential relationships and then
applying theoretical frameworks to interpret and validate these
findings. Such an approach could provide a more nuanced under-
standing of cognitive resilience and its underlying mechanisms. Finally,
while some of the effect sizes in this study were small, small effects
can accumulate over time and are meaningful at the population level
(Funder & Ozer, 2019).
This study provides a comprehensive characterization of the
predictors of cognitive resilience during adolescence. Using a large,
diverse sample of children from low‐income households, we found
that several factors across all three domains of individual, family, and
neighborhood characteristics predict both fluid and crystallized
cognitive resilience. Our findings highlight the importance of a
multifaceted approach to promoting cognitive resilience among
children from low‐income households. Future intervention work
should examine the role of modifiable factors such as physical activity
opportunities, access to and encouragement of involvement in
extracurricular activities, support for women during pregnancy and
postpartum, and high‐quality early childhood education. These ef-
forts can help to improve the long‐term outcomes for children from
low‐income households.
PREDICTORS OF COGNITIVE RESILIENCE IN LOW‐INCOME YOUTH
-
11 of 14
AUTHOR CONTRIBUTION
Divyangana Rakesh: Conceptualization (lead); Funding acquisition
(lead); Methodology (equal); Project administration (lead); Writing –
Original Draft (lead); Writing – Review & Editing (equal). Ekaterina
Sadikova: Methodology (equal); Formal analysis (lead); Writing –
Original Draft (supporting); Writing – Review & Editing (equal). Katie
McLaughlin: Supervision (lead); Writing – Review & Editing (equal).
ACKNOWLEDGMENTS
Data used in the preparation of this article were obtained from the
ABCD Study (https://abcdstudy.org) held in the NDA. This is a
multisite, longitudinal study designed to recruit more than 10,000
children ages 9–10 years old and follow them over 10 years into early
adulthood. The ABCD study is supported by the National Institutes of
Health and additional federal partners under award numbers
U01DA041048, U01DA050989, U01DA051016, U01DA041022,
U01DA051018, U01DA051037, U01DA050987, U01DA041174,
U01DA041106, U01DA041117, U01DA041028, U01DA041134,
U01DA050988, U01DA051039, U01DA041156, U01DA041025,
U01DA041120, U01DA051038, U01DA041148, U01DA041093,
U01DA041089, U24DA041123 and U24DA041147. A full list of
supporters is available at https://abcdstudy.org/federal‐partners.
html. A listing of participating sites and a complete listing of the study
investigators can be found at https://abcdstudy.org/consortium_
members/. Adolescent Brain Cognitive Development consortium in-
vestigators designed and implemented the study and/or provided
data but did not necessarily participate in the analysis or writing of
this report. The views expressed in this manuscript are those of the
authors and do not necessarily reflect the official views of the Na-
tional Institutes of Health, the Department of Health and Human
Services, the US federal government or ABCD consortium in-
vestigators. This work was conducted with support from the Harvard
Mind Brain and Behavior Interfaculty Initiative (granted to DR). KM
was supported by R01‐MH106482 and R37‐MH119194.
CONFLICT OF INTEREST STATEMENT
The authors report no biomedical financial interests or potential
conflicts of interest.
DATA AVAILABILITY STATEMENT
The Adolescent Brain Cognitive Development (ABCD) Study data is
available for qualified researchers who have an approved Data Use
Certificate (DUC) from the National Institute of Mental Health
(NIMH) Data Archive (NDA).
ETHICAL CONSIDERATIONS
The need for ethics approval was waived by The University of Cali-
fornia, Los Angeles, institutional review board (IRB) stating that
secondary analyses using the publicly released ABCD Study data are
not human subjects research and therefore do not require their own
approval. The ABCD Study received their own central IRB approval.
All guidelines pertaining to the Declaration of Helsinki were adhered
to. Caregivers provided written informed consent and children pro-
vided assent for participation in the study.
ORCID
Divyangana Rakesh
https://orcid.org/0000-0002-8529-2086
Ekaterina Sadikova https://orcid.org/0000-0002-2050-9475
Katie McLaughlin https://orcid.org/0000-0002-1362-2410
REFERENCES
Akshoomoff, N., Beaumont, J. L., Bauer, P. J., Dikmen, S. S., Gershon, R. C.,
Mungas, D., Slotkin, J., Tulsky, D., Weintraub, S., Zelazo, P. D., &
Heaton, R. K. (2013). National Institutes of health toolbox cognition
battery (NIH toolbox CB): Validation for children between 3 and 15
years: VIII. NIH toolbox cognition battery (CB): Composite scores of
crystallized, fluid, and overall cognition. Monographs of the Society for
Research in Child Development,78(4), 119–132. https://doi.org/10.
1111/mono.12038
ALICE in Focus. (2022). ALICE in focus: Children. National Reaearch Brief.
https://www.unitedforalice.org/committees/focus‐children
Annie E. Casey Foundation. (2014). The 2014 KIDS COUNT data book—
the Annie E. Casey foundation. Retrieved from https://www.aecf.
org/resources/2019‐kids‐count‐data‐book/
Best, J. R., Miller, P. H., & Naglieri, J. A. (2011). Relations between exec-
utive function and academic achievement from ages 5 to 17 in a
large, representative national sample. Learning and Individual Differ-
ences,21(4), 327–336. https://doi.org/10.1016/j.lindif.2011.01.007
Brody, G. H., Yu, T., Nusslock, R., Barton, A. W., Miller, G. E., Chen, E., Holmes,
C., McCormick, M., & Sweet, L. H. (2019). The protective effects of
supportive parenting on the relationship between adolescent poverty
and resting‐state functional brain connectivity during adulthood.
Psychological Science,30(7), 1040–1049. https://doi.org/10.1177/
0956797619847989
Bronfenbrenner, U. (1994). Ecological models of human development.
Readings on the Development of Children,2(1), 37–43.
Burton, L. M., & Robin, R. L. (2000). In the mix, yet on the margins: The place
of families in urban neighborhood and child development research.
Journal of Marriage and Family,62(4), 1114–1135. https://doi.org/10.
1111/J.1741‐3737.2000.01114.X
Buuren, S. van, & Groothuis‐Oudshoorn, K. (2011). mice: Multivariate
imputation by chained equations in R. Journal of Statistical Software,
45(3), 1–67. https://doi.org/10.18637/jss.v045.i03
Cattell, R. B. (1963). Theory of fluid and crystallized intelligence: A critical
experiment. Journal of Educational Psychology,54(1), 1–22. https://
doi.org/10.1037/h0046743
Cattell, R. B. (1987). Intelligence: Its structure, growth and action (pp. xxii–
694). North‐Holland.
Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system.
In Proceedings of the 22nd ACM SIGKDD international conference on
knowledge discovery and data mining (pp. 785–794). https://doi.org/
10.1145/2939672.2939785
Chmielewski, A. K. (2019). The global increase in the socioeconomic
achievement gap, 1964 to 2015. American Sociological Review,84(3),
517–544. https://doi.org/10.1177/0003122419847165
Cooper, K., & Stewart, K. (2013). Does money affect children’s outcomes? A
systematic review. Joseph Rowtree Foundation.
Crouch, E., Probst, J. C., Radcliff, E., Bennett, K. J., & McKinney, S. H. (2019).
Prevalence of adverse childhood experiences (ACEs) among US chil-
dren. Child Abuse and Neglect,92, 209–218. https://doi.org/10.1016/j.
chiabu.2019.04.010
Davis‐Kean, P. E. (2005). The influence of parent education and family in-
come on child achievement: The indirect role of parental expectations
and the home environment. Journal of Family Psychology: JFP: Journal of
the Division of Family Psychology of the American Psychological Associa-
tion,19(2), 294–304. https://doi.org/10.1037/0893‐3200.19.2.294
DeJoseph, M. L., Ellwood‐Lowe, M. E., Miller‐Cotto, D., Silverman, D.,
Shannon, K. A., Reyes, G., Rakesh, D., & Frankenhuis, W. E. (2024). The
promise and pitfalls of a strength‐based approach to child poverty and
neurocognitive development: Implications for policy. Developmental
Cognitive Neuroscience,66, 101375. https://doi.org/10.1016/j.dcn.
2024.101375
Distefano, R., Palmer, A. R., Kalstabakken, A. W., Hillyer, C. K., Seiwert, M. J.,
Zelazo, P. D., Carlson, S. M., & Masten, A. S. (2023). Predictive validity
of the NIH toolbox executive function measures with developmental
extensions from early childhood to third grade achievement.
12 of 14
-
RAKESH
ET AL.
Developmental Neuropsychology,48(8), 373–386. https://doi.org/10.
1080/87565641.2023.2286353
Ellis, B. J., Sheridan, M. A., Belsky, J., & McLaughlin, K. A. (2022). Why and
how does early adversity influence development? Toward an inte-
grated model of dimensions of environmental experience. Develop-
ment and Psychopathology,34(2), 447–471. https://doi.org/10.1017/
S0954579421001838
Ellwood‐Lowe, M. E., Irving, C. N., & Bunge, S. A. (2022). Exploring neural
correlates of behavioral and academic resilience among children in
poverty. Developmental Cognitive Neuroscience,54, 101090. https://
doi.org/10.1016/j.dcn.2022.101090
Foster, E. M., & Marcus Jenkins, J. V. (2017). Does participation in music
and performing arts influence child development? American Educa-
tional Research Journal,54(3), 399–443. https://doi.org/10.3102/
0002831217701830
Funder, D. C., & Ozer, D. J. (2019). Evaluating effect size in psychological
research: Sense and nonsense. Advances in Methods and Practices in
Psychological Science,2(2), 156–168. https://doi.org/10.1177/
2515245919847202
Garavan, H., Bartsch, H., Conway, K., Decastro, A., Goldstein, R. Z.,
Heeringa, S., Jernigan, T., Potter, A., Thompson, W., & Zahs, D.
(2018). Recruiting the ABCD sample: Design considerations and
procedures. Developmental Cognitive Neuroscience,32, 16–22. https://
doi.org/10.1016/j.dcn.2018.04.004
Goetschius, L. G., McLoyd, V. C., Hein, T. C., Mitchell, C., Hyde, L. W., &
Monk, C. S. (2023). School connectedness as a protective factor
against childhood exposure to violence and social deprivation: A
longitudinal study of adaptive and maladaptive outcomes. Develop-
ment and Psychopathology,35(3), 1219–1234. https://doi.org/10.
1017/S0954579421001140
Graham, E. K., James, B. D., Jackson, K. L., Willroth, E. C., Boyle, P., Wilson,
R., Bennett, D. A., & Mroczek, D. K. (2021). Associations between
personality traits and cognitive resilience in older adults. The Jour-
nals of Gerontology: Serie Bibliographique,76(1), 6–19. https://doi.org/
10.1093/geronb/gbaa135
Greenman, E., Bodovski, K., & Reed, K. (2011). Neighborhood character-
istics, parental practices and children’s math achievement in
elementary school. Social Science Research,40(5), 1434–1444.
https://doi.org/10.1016/J.SSRESEARCH.2011.04.007
Gruijters, R. J., Raabe, I. J., & Hübner, N. (2024). Socio‐emotional skills and
the socioeconomic achievement gap. Sociology of Education,97(2),
120–147. https://doi.org/10.1177/00380407231216424
Haft, S. L., Myers, C. A., & Hoeft, F. (2016). Socio‐emotional and cognitive
resilience in children with reading disabilities. Current Opinion in
Behavioral Sciences,10, 133–141. https://doi.org/10.1016/j.cobeha.
2016.06.005
Harmelen, A.‐L. van, Kievit, R. A., Ioannidis, K., Neufeld, S., Jones, P. B.,
Bullmore, E., Dolan, R., Consortium, T. N., Fonagy, P., & Goodyer, I.
(2017). Adolescent friendships predict later resilient functioning
across psychosocial domains in a healthy community cohort. Psy-
chological Medicine,47(13), 2312–2322. https://doi.org/10.1017/
S0033291717000836
Haverkamp, B. F., Wiersma, R., Vertessen, K., van Ewijk, H., Oosterlaan,
J., & Hartman, E. (2020). Effects of physical activity interventions on
cognitive outcomes and academic performance in adolescents and
young adults: A meta‐analysis. Journal of Sports Sciences,38(23),
2637–2660. https://doi.org/10.1080/02640414.2020.1794763
Heaton, R. K., Akshoomoff, N., Tulsky, D., Mungas, D., Weintraub, S.,
Dikmen, S., Beaumont, J., Casaletto, K. B., Conway, K., Slotkin, J., &
Gershon, R. (2014). Reliability and validity of composite scores from
the NIH Toolbox Cognition Battery in adults. Journal of the Interna-
tional Neuropsychological Society: JINS,20(6), 588–598. https://doi.
org/10.1017/S1355617714000241
Hinnant, J. B., El‐Sheikh, M., Keiley, M., & Buckhalt, J. A. (2013). Marital
conflict, allostatic load, and the development of children’s fluid
cognitive performance. Child Development,84(6), 2003–2014.
https://doi.org/10.1111/cdev.12103
Horta, B. L., Loret de Mola, C., & Victora, C. G. (2015). Breastfeeding and
intelligence: A systematic review and meta‐analysis. Acta Paediatrica,
104(S467), 14–19. https://doi.org/10.1111/apa.13139
Kirlic, N., Colaizzi, J. M., Cosgrove, K. T., Cohen, Z. P., Yeh, H.‐W., Breslin,
F., Morris, A. S., Aupperle, R. L., Singh, M. K., & Paulus, M. P. (2021).
Extracurricular activities, screen media activity, and sleep may
Be modifiable factors related to children’s cognitive functioning:
Evidence from the ABCD Study®. Child Development,92(5), 2035–
2052. https://doi.org/10.1111/cdev.13578
Klebanov, P. K., Brooks‐Gunn, J., & Duncan, G. J. (1994). Does neighbor-
hood and family poverty affect mothers’ parenting, mental health,
and social support? Journal of Marriage and Family,56(2), 441. https://
doi.org/10.2307/353111
Kong, K. (2020). Academic resilience of pupils from low socioeconomic
backgrounds. The Journal of Behavioral Science,15(2). Article 2.
Lawson, G. M., Hook, C. J., & Farah, M. J. (2018). A meta‐analysis of the
relationship between socioeconomic status and executive function
performance among children. Developmental Science,21(2), e12529.
https://doi.org/10.1111/DESC.12529
Lloyd, J. E. V., Li, L., & Hertzman, C. (2010). Early experiences matter:
Lasting effect of concentrated disadvantage on children’s language
and cognitive outcomes. Health and Place,16(2), 371–380. https://
doi.org/10.1016/J.HEALTHPLACE.2009.11.009
Luciana, M., Bjork, J. M., Nagel, B. J., Barch, D. M., Gonzalez, R., Nixon,
S. J., & Banich, M. T. (2018). Adolescent neurocognitive development
and impacts of substance use: Overview of the adolescent brain
cognitive development (ABCD) baseline neurocognition battery.
Developmental Cognitive Neuroscience,32, 67–79. https://doi.org/10.
1016/j.dcn.2018.02.006
Lund, J. I., Toombs, E., Radford, A., Boles, K., & Mushquash, C. (2020).
Adverse childhood experiences and executive function difficulties in
children: A systematic review. Child Abuse and Neglect,106, 104485.
https://doi.org/10.1016/j.chiabu.2020.104485
Lundberg, S., & Lee, S.‐I. (2017). A unified approach to interpreting model
predictions (arXiv:1705.07874). arXiv.https://doi.org/10.48550/
arXiv.1705.07874
Luthar, S. S., & Cicchetti, D. (2000). The construct of resilience: Implications
for interventions and social policies. Development and Psychopathology,
12(4), 857–885. https://doi.org/10.1017/S0954579400004156
Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resilience:
A critical evaluation and guidelines for future work. Child Develop-
ment,71(3), 543–562. https://doi.org/10.1111/1467‐8624.00164
Masten, A. S. (2014). Global perspectives on resilience in children and youth.
Child Development,85(1), 6–20. https://doi.org/10.1111/cdev.12205
Masten, A. S., Lucke, C. M., Nelson, K. M., & Stallworthy, I. C. (2021).
Resilience in development and psychopathology: Multisystem per-
spectives. Annual Review of Clinical Psychology,17(1), 521–549.
https://doi.org/10.1146/Annurev‐Clinpsy‐081219‐120307
McLaughlin, M., & Rank, M. R. (2018). Estimating the economic cost of
childhood poverty in the United States. Social Work Research,42(2),
73–83. https://doi.org/10.1093/swr/svy007
McLoyd, V. C. (1998). Socioeconomic disadvantage and child develop-
ment. American Psychologist,53(2), 185–204. https://doi.org/10.
1037/0003‐066X.53.2.185
Noble, K. G., McCandliss, B. D., & Farah, M. J. (2007). Socioeconomic
gradients predict individual differences in neurocognitive abilities.
Developmental Science,10(4), 464–480. https://doi.org/10.1111/j.
1467‐7687.2007.00600.x
O’Malley, M., Voight, A., Renshaw, T. L., & Eklund, K. (2015). School climate,
family structure, and academic achievement: A study of moderation
effects. School Psychology Quarterly,30(1), 142–157. https://doi.org/
10.1037/spq0000076
Papini, S., Norman, S. B., Campbell‐Sills, L., Sun, X., He, F., Kessler, R. C.,
Ursano, R. J., Jain, S., & Stein, M. B. (2023). Development and validation
of a machine learning prediction model of posttraumatic stress dis-
order after military deployment. JAMA Network Open,6(6), e2321273.
https://doi.org/10.1001/jamanetworkopen.2023.21273
Piccolo, L. R., Merz, E. C., & Noble, K. G. (2019). School climate is associated
with cortical thickness and executive function in children and ado-
lescents. Developmental Science,22(1), 1–11. https://doi.org/10.1111/
desc.12719
Rakesh, D., Lee, P. A., Gaikwad, A., & McLaughlin, K. A. (2024). Associations
of socioeconomic status with cognitive function, language ability, and
PREDICTORS OF COGNITIVE RESILIENCE IN LOW‐INCOME YOUTH
-
13 of 14
academic achievement in youth: A systematic review of mechanisms
and protective factors. Journal of Child Psychology and Psychiatry.
Portico. https://doi.org/10.1111/jcpp.14082
Rakesh, D., Seguin, C., Zalesky, A., Cropley, V., & Whittle, S. (2021). Asso-
ciations between neighborhood disadvantage, resting‐state func-
tional connectivity, and behavior in the adolescent brain cognitive
development study: The moderating role of positive family and school
environments. Biological Psychiatry: Cognitive Neuroscience and Neuro-
imaging,6(9), 877–886. https://doi.org/10.1016/j.bpsc.2021.03.008
Rakesh, D., & Whittle, S. (2021). Socioeconomic status and the developing
brain – a systematic review of neuroimaging findings in youth.
Neuroscience & Biobehavioral Reviews,130, 379–407. https://doi.org/
10.1016/j.neubiorev.2021.08.027
Rakesh, D., Whittle, S., Sheridan, M. A., & McLaughlin, K. A. (2023).
Childhood socioeconomic status and the pace of structural neuro-
development: Accelerated, delayed, or simply different? Trends in
Cognitive Sciences,27(9), 833–851. https://doi.org/10.1016/j.tics.
2023.03.011
Rakesh, D., Zalesky, A., & Whittle, S. (2021). Similar but distinct – Effects
of different socioeconomic indicators on resting state functional
connectivity: Findings from the Adolescent Brain Cognitive Devel-
opment (ABCD) Study®. Developmental Cognitive Neuroscience,51,
101005. https://doi.org/10.1016/j.dcn.2021.101005
Rakesh, D., Zalesky, A., & Whittle, S. (2022). Assessment of parent income
and education, neighborhood disadvantage, and child brain struc-
ture. JAMA Network Open,5(8), e2226208. https://doi.org/10.1001/
JAMANETWORKOPEN.2022.26208
Raniti, M., Rakesh, D., Patton, G. C., & Sawyer, S. M. (2022). The role of
school connectedness in the prevention of youth depression and
anxiety: A systematic review with youth consultation. BMC Public
Health,22(1), 2152. https://doi.org/10.1186/s12889‐022‐14364‐6
Raver, C. C., Blair, C., & Willoughby, M. (2013). Poverty as a predictor of
4‐year‐olds’ executive function: New perspectives on models of
differential susceptibility. Developmental Psychology,49(2), 292–304.
https://doi.org/10.1037/a0028343
Russell, A. E., Ford, T., Williams, R., & Russell, G. (2016). The association
between socioeconomic disadvantage and attention deficit/hyper-
activity disorder (adhd): A systematic review. Child Psychiatry and
Human Development,47(3), 440–458. https://doi.org/10.1007/
s10578‐015‐0578‐3
Rutter, M. (2006). Implications of resilience concepts for scientific un-
derstanding. Annals of the New York Academy of Sciences,1094, 1–12.
https://doi.org/10.1196/annals.1376.002
Sattler, K., & Gershoff, E. (2019). Thresholds of resilience and within‐and
cross‐domain academic achievement among children in poverty.
Early Childhood Research Quarterly,46, 87–96. https://doi.org/10.
1016/j.ecresq.2018.04.003
Shaw, C., Wu, Y., Zimmerman, S. C., Hayes‐Larson, E., Belin, T. R., Power,
M. C., Glymour, M. M., & Mayeda, E. R. (2023). Comparison of
imputation strategies for incomplete longitudinal data in life‐course
epidemiology. American Journal of Epidemiology,192(12), 2075–2084.
https://doi.org/10.1093/aje/kwad139
Shenkin, S. D., Starr, J. M., & Deary, I. J. (2004). Birth weight and
cognitive ability in childhood: A systematic review. Psychological
Bulletin,130(6), 989–1013. https://doi.org/10.1037/0033‐2909.130.
6.989
Shumow, L., & Lomax, R. (2009). Parental efficacy: Predictor of parenting
behavior and adolescent outcomes. Parenting,2(2), 127–150. https://
doi.org/10.1207/S15327922PAR0202_03
Stumper, A., Mac Giollabhui, N., Abramson, L. Y., & Alloy, L. B. (2020). Early
pubertal timing mediates the association between low socioeconomic
status and poor attention and executive functioning in a diverse
community sample of adolescents. Journal of Youth and Adolescence,
49(7), 1420–1432. https://doi.org/10.1007/s10964‐020‐01198‐x
Sylvers, D. L., Hicken, M., Esposito, M., Manly, J., Judd, S., & Clarke, P.
(2022). Walkable neighborhoods and cognition: Implications for the
design of health promoting communities. Journal of Aging and Health,
34(6–8), 893–904. https://doi.org/10.1177/08982643221075509
Tikhomirova, T., Malykh, A., & Malykh, S. (2020). Predicting academic
achievement with cognitive abilities: Cross‐sectional study across
school education. Behavioral Sciences,10(10), 158. https://doi.org/10.
3390/bs10100158
Valcan, D. S., Davis, H., & Pino‐Pasternak, D. (2018). Parental behaviours
predicting early childhood executive functions: A meta‐analysis.
Educational Psychology Review,30(3), 607–649. https://doi.org/10.
1007/s10648‐017‐9411‐9
Vella‐Brodrick, D. A., & Gilowska, K. (2022). Effects of nature (greenspace)
on cognitive functioning in school children and adolescents: A sys-
tematic review. Educational Psychology Review,34(3), 1217–1254.
https://doi.org/10.1007/s10648‐022‐09658‐5
Whittle, S., Vijayakumar, N., Simmons, J. G., Dennison, M., Schwartz, O.,
Pantelis, C., Sheeber, L., Byrne, M. L., & Allen, N. B. (2017). Role of
positive parenting in the association between neighborhood social
disadvantage and brain development across adolescence. JAMA
Psychiatry,74(8), 824–832. https://doi.org/10.1001/jamapsychiatry.
2017.1558
Willroth, E. C., James, B. D., Graham, E. K., Kapasi, A., Bennett, D. A., &
Mroczek, D. K. (2023). Well‐being and cognitive resilience to
dementia‐related neuropathology. Psychological Science,34(3), 283–
297. https://doi.org/10.1177/09567976221119828
Yan, Y., & Gai, X. (2022). High achievers from low family socioeconomic
status families: Protective factors for academically resilient stu-
dents. International Journal of Environmental Research and Public
Health,19(23), 15882. Article 23. https://doi.org/10.3390/
ijerph192315882
Zhang, L., Rakesh, D., Cropley, V., & Whittle, S. (2023). Neurobiological
correlates of resilience during childhood and adolescence – a sys-
tematic review. Clinical Psychology Review,105, 102333. https://doi.
org/10.1016/j.cpr.2023.102333
SUPPORTING INFORMATION
Additional supporting information can be found online in the Sup-
porting Information section at the end of this article.
How to cite this article: Rakesh, D., Sadikova, E., &
McLaughlin, K. (2024). Beyond the income‐achievement gap:
The role of individual, family, and environmental factors in
cognitive resilience among low‐income youth. JCPP Advances,
e12297. https://doi.org/10.1002/jcv2.12297
14 of 14
-
RAKESH
ET AL.
