Effects of Child Development Accounts on Adolescent Behavior Problems: Evidence from a Longitudinal, Randomized Policy Experiment

Abstract

Theory and empirical examination have shown ways in which households’ asset building for children may affect child well-being, including behavioral and emotional health. Previous research found that Child Development Accounts (CDAs), a universal and lifelong asset-building policy designed to encourage society and families to accumulate assets for children, have positive effects on social-emotional development at around 4 years of age. Using data from a unique longitudinal experiment of the CDA policy in the United States, this study examined the impacts of CDAs on adolescent behavior problems. Adolescent behavior problems were indicated by eight items from a 28-item Behavior Problems Index, focusing on children’s anxiety and disobedience. In the pre-COVID sample (n = 676), results from the regression analysis show that the sum score of child behavior problems is about 0.12 standard deviations higher than that of counterparts in the control group, indicating less frequent behavior problems among children in the treatment group. Regression analyses on two latent measures of anxiety and disobedience showed that CDAs primarily affected children’s anxiety. However, there were no statistically significant differences in the full sample (N = 1712); this could be related to the data-collection disruptions caused by the COVID pandemic. This study provides the first longitudinal follow-up evidence on the effects of CDAs on adolescent behavior problems. The findings show that universal asset building for adolescents reduces behavior problems. Policy implications are discussed.

1. Introduction

Adolescence is a transformative phase characterized by a multitude of physical, cognitive, emotional, and social changes. Navigating these changes can be challenging, and a substantial proportion of adolescents experience behavioral and emotional difficulties. These challenges—such as persistent anxiety, disobedience, or social withdrawal—are often viewed as early indicators of psychological distress, with immediate and long-term consequences for health and development (Adams 2005; Christiansen et al. 2021).

Recent data highlight the urgency of addressing adolescent mental health. In the United States, four in ten (40%) of high school students had persistent feelings of sadness or hopelessness, and two in ten (20%) reported having seriously considered attempting suicide in the prior year (Centers for Disease Control and Prevention 2024). Such behavioral and emotional difficulties during adolescence are linked to poor academic performance and later mental health concerns in adulthood (Agnafors et al. 2021; Kim-Cohen et al. 2003; McGue and Iacono 2005; Reef et al. 2009).

Family socioeconomic status (SES) is a well-established social determinant of youth behavioral and emotional health. While early studies focused on income-based measures, recent research has emphasized the role of structural and contextual factors, including economic stability, caregiver well-being, and neighborhood disadvantage, in shaping mental health outcomes (Arcaya et al. 2016). For example, economic hardship can increase parental stress, reduce parental engagement, and destabilize home environments—each of which is associated with adolescent behavioral risk (Low and Mounts 2022; Peverill et al. 2021).

To move beyond traditional SES indicators, researchers have increasingly focused on household assets—such as savings, homeownership, and long-term financial planning—as critical, future-oriented resources that buffer families from economic shocks and support healthy child development (Ansong et al. 2024; Gibson-Davis and Hill 2021). According to asset-effects theory (Sherraden 1991), assets influence child well-being not only materially but also psychologically—by promoting stability, long-term orientation, and a sense of security.

Family investment theory offers a complementary lens, positing that parents with greater economic resources are better able to invest in enriching home environments and engage in future-oriented parenting practices that promote socioemotional development (Becker 1981; Yeung et al. 2002). In the context of adolescent behavior, increased household assets may reduce economic stress, foster future planning, and enhance parent-adolescent relationships—thus supporting behavioral health. Grinstein-Weiss et al. (2014) proposed four pathways through which parental assets may positively impact children’s well-being. These pathways include assets serving as financial resources to mitigate financial crises and hardships and assets providing a foundation for fostering positive parent–youth interactions. In the context of adolescent behavioral health, it is plausible that increased family assets allow for investments in enhanced home-learning environments and future-oriented, goal-directed interactions between parents and youth, thereby potentially improving adolescents’ behavioral competencies. Families with greater assets may also be more resilient when facing financial shocks, as they have resources to buffer income loss. Such resources could enable parents to navigate family tensions that might otherwise negatively impact child well-being (Lerman and McKernan 2008).

Child Development Accounts (CDAs) are an institutional intervention that supports asset building from early childhood. These accounts are designed to promote long-term saving and investment in children’s development, particularly in education. Evidence from experimental studies has shown that CDAs are associated with positive social-emotional outcomes in early childhood (Huang et al. 2014), confidence in achieving educational plans (Curley et al. 2010), and parenting attitudes and involvement (Huang et al. 2021).

However, little is known about the longer-term effects of CDAs on adolescent behavioral health, especially during the critical transition from childhood to adolescence. This study represents the first attempt to examine whether holding assets in dedicated accounts for children affects the risk of adolescent behavior problems. We hypothesized that holding assets in a dedicated account for children would reduce the risk of behavior problems in adolescents. We tested the hypothesis through analyses of data from a unique longitudinal experiment on CDA policy in the United States.

2. Methods

2.1. The SEED for Oklahoma Kids Experiment

The SEED for Oklahoma Kids (SEED OK) experiment is a large-scale test of a statewide CDA policy. The experiment’s probability sample was randomly drawn from infants in Oklahoma at birth (Huang et al. 2021). The sampling frame was the population of Oklahoma infants born in two 3-month periods in 2007, with oversampling of African Americans, American Indians, and Hispanics. The randomized experiment’s design (Sherraden et al. 2018) allows researchers to attribute causality of observed impacts to the CDA policy intervention and then to generalize those results to the full cohort in the Oklahoma state population.

For each child in the treatment group, the Oklahoma treasurer’s office automatically opened a state-owned account with a $1000 deposit through the Oklahoma 529 College Savings Plan (OK 529). In addition, SEED OK offered two financial incentives to encourage treatment families to open their own OK 529 accounts for the children and make deposits with their own money. The first financial incentive was the offer of a $100 initial contribution to OK 529 accounts opened by families for treatment children by a given date. The second financial incentive was the provision of savings matches for 4 years (2008–2011) to low- and moderate-income treatment families that made deposits in the individual-owned accounts. Account statements have been sent to treatment families in each calendar quarter between 2008 and 2019 and in each calendar year since. Treatment families also have been sent other program materials (e.g., letters and postcards). SEED OK was a low-resource intervention; no other measures directly touched the treatment mothers. Families in the control group were eligible to open an OK 529 account for their child, but they did not receive the CDA intervention components. More details on the experiment can be found in prior SEED OK research (Huang et al. 2014). Figure 1 presents SEED OK’s randomization and intervention components. This study was approved by the Institutional Review Board (IRB) of our university and was conducted in accordance with the ethical standards outlined in the Helsinki Declaration on human subjects testing (World Medical Association 2022).

2.2. Data and Sample

In 2007, primary caregivers of 2704 randomly selected children agreed to participate in the SEED OK experiment. (Because 99% of the caregivers were mothers, we refer to all of them as mothers in this paper.) After completing the Wave 1 (baseline) survey between fall 2007 and spring 2008, these mothers were randomly assigned to the treatment (n = 1358) or control group (n = 1346; Marks et al. 2008). In spring 2011, 84% of the mothers completed the Wave 2 survey (n = 2259; 1149 in the treatment group and 1110 in the control group). The Wave 3 survey began in January 2020, when the children were between 12 and 13 years old. By July 2020, 67% of the mothers completed the Wave 3 survey (n = 1799; 921 in the treatment group and 878 in the control group). This study uses data from the Wave 1 and Wave 3 surveys of the SEED OK experiment.

After data collection began in Wave 3, the COVID pandemic disrupted daily routines, activities, and arrangements for everyone involved in the experiment. These disruptions may have affected the SEED OK treatment’s links to outcomes for SEED OK children and their families. The pandemic also forced several changes in survey-data collection, including a shift from a phone survey to an online one and the cancellation of in-person field visits. Because of the pandemic’s known and potential effects on the lives of SEED OK participants, the research methodology, and the accuracy of measurement, we analyzed data from the pre-COVID sample alone (n = 707; 369 in the treatment group and 338 in the control group) and from the full sample (n = 1799). Observations with missing values were removed (31 from the pre-COVID sample and 87 from the full sample). Data were collected from the pre-COVID sample prior to 1 April 2020. Because COVID may have affected both SEED OK’s impacts and our ability to measure any impacts, we believe that the results from the pre-COVID sample are more reliable and generalizable than those from the full sample.

2.3. Outcome and Focal Independent Variables

The dependent variables that measure adolescents’ behavior problems were based on eight survey items derived from the 28-item Behavior Problems Index (BPI; Peterson and Zill 1986):

• “The child has sudden changes in mood or feeling.”
• “The child feels or complains that no one loves him/her.”
• “The child is too fearful or anxious.”
• “The child is unhappy, sad, or depressed.”
• “The child feels worthless or inferior.”
• “The child is disobedient at school.”
• “The child has trouble getting along with other children.”
• “The child is disobedient at home.”

Each item had three possible responses (1, often true; 2, sometimes true; and 3, not true). The first five items came from the Anxious/Depression subscale of the BPI, which measures internalized behavior problems directed inwardly toward oneself. The three others came from the Antisocial, Headstrong, and Peer Conflicts subscales, respectively, focusing on children’s disobedience and antisocial behaviors; they indicate externalizing behaviors directed outwardly to others or the environment. A higher value indicates less frequent behavior problems. Due to the length of the survey, we were not able to include all 28 BPI questions.

The focal independent variable was the CDA treatment status (1, treatment; 0, control). We controlled for Wave 1 demographic and socioeconomic characteristics of the children, mothers, and households.

2.4. Statistical Analysis

We developed two measures of children’s behavior problems. These were based on eight items from the SEED OK Wave 3 survey. The first measure was a sum scale (0–8) created by using all eight items and dichotomizing responses according to the BPI scoring guide (1, not true; 0, often true or sometimes true). Higher scores represent more-desired child-behavior outcomes—that is, less frequent behavior problems.

The second measure was created by applying a confirmatory factor analysis (CFA) to generate latent measures of anxiety and disobedience. We used CFA because only a small proportion of the 28 BPI items were included in our survey, and it was not clear how well these variables would hold together as adequate indicators of children’s behavior problems. Results from CFA would demonstrate how well the measured variables represented the constructs that these items were intended to measure. Through regression analyses without and with controls for Wave 1 demographic and socioeconomic characteristics (see

Table 1. Characteristics of the Pre-COVID Sample (n = 676).

Characteristic	Control (n = 328)	Treatment (n = 348)	Possible Responses
Child
Male (%)	50.02	54.32	1 = Male 0 = Female
Race (%)
Non-Hispanic White	69.45	70.29	1 = Non-Hispanic White 2 = Non-Hispanic African Americans 3 = Non-Hispanic American Indian 4 = Non-Hispanic Asian American and Pacific Islander 5 = Hispanic
Non-Hispanic African American	7.26	8.19
Non-Hispanic American Indian	8.28	9.45
Non-Hispanic Asian American and Pacific Islander	1.98	2.05
Hispanic	13.04	10.02
Child’s age, M (SD), by year	12.48 (0.19)	12.48 (0.18)
Mother
Age, M (SD), by year	26.38 (5.64)	26.67 (5.50)
Education (%)
Below high school	17.15	14.75	1 = Below high school diploma 2 = High school diploma or general equivalency diploma 3 = Some college 4 = 4-year college or above
High school	32.09	28.07
Some college	21.64	25.77
4-year college or above	28.22	31.42
Health status, M (SD)	4.16 (0.91)	4.20 (0.88)	1 (poor)–5 (excellent)
Marital status (% married)	67.58	67.53	1 = Married 2 = Other
Employment status (% working)	56.29	56.60	1 = Working 0 = Not working
Household
Household size, M (SD)	2.97 (1.10)	3.20 (1.31)
Save monthly (% yes)	70.49	64.56
Parenting attitude index, M (SD)	11.02 (1.38)	11.12 (1.40)	6–14 (higher score indicates positive attitudes)
Homeownership (% yes)	54.77	54.50	1 = Yes 0 = No
TANF participation (% yes)	7.35	7.84	1 = Yes 0 = No
SNAP participation (% yes)	32.33	30.18	1 = Yes 0 = No
Income-to-needs ratio, M (SD)	241.54 (262.31)	233.86 (239.04)
English as primary language at home (% yes) *	90.90	95.21	1 = Yes 0 = No
Geographic areas (%) *
Metropolitan area	62.09	71.69	1 = Metropolitan area 2 = Micropolitan area 3 = Other
Micropolitan area	22.75	16.84
Other	15.16	11.47

Note. Results are weighted. Two-tailed statistical tests. TANF = Temporary Assistance for Needy Families; SNAP = Supplemental Nutrition Assistance Program. * p < 0.10.

), we compared the differences between the treatment and control groups in the sum scale and the latent measures of children’s anxiety and disobedience. We conducted these two sets of analyses, one with data from the pre-COVID sample (n = 676) and another with data from the full Wave 3 sample (N = 1712). Consistent results emerged from regression analyses without and with controls. We report results from regressions estimated with controls.

All analyses were conducted using Stata version 18. We applied sampling weights in all analyses. Based on our directional hypotheses, we used one-tailed tests and 0.10 as the significance level to test the CDA effects for the two sets of analyses.

3. Results

3.1. Sample Descriptions

Table 1 presents Wave 1 characteristics for children, mothers, and households in the pre-COVID sample (n = 676). The treatment and control groups did not differ significantly, except that treatment mothers were more likely to have English as their primary language at home and to live in a metropolitan area. Similar descriptive results (not reported) hold for the full sample (N = 1712). This suggests that the treatment and control groups generated through randomization at Wave 1 remained comparable after 13 years.

3.2. Measures of Children’s Behavior Problems

The first two columns of

Table 2. Children’s behavior problems: mean scores of the pre-COVID sample (n = 676).

Measure	Treatment (n = 348)	Control (n = 328)	Treatment-Control Difference	CFA Loading on 2 Latent Measures
Anxiety
1. Sudden mood changes	2.46 (0.61)	2.42 (0.61)	+0.04	0.59 (0.06) ***
2. Complaining no love	2.89 (0.37)	2.83 (0.43)	+0.06 *	0.87 (0.05) ***
3. Too fearful	2.67 (0.57)	2.57 (0.58)	+0.10 **	0.44 (0.07) ***
4. Feel unhappy or depressed	2.72 (0.52)	2.71 (0.46)	+0.01	0.76 (0.06) ***
5. Feel worthless	2.82 (0.43)	2.75 (0.45)	+0.07 *	0.77 (0.05) ***
Disobedience
6. Disobedient at school	2.81 (0.48)	2.82 (0.44)	−0.01	0.65 (0.08) ***
7. Not getting along with others	2.81 (0.45)	2.80 (0.42)	+0.01	0.79 (0.07) ***
8. Disobedient at home	2.54 (0.60)	2.59 (0.52)	−0.05	0.67 (0.08) ***
Sum score after dichotomization (0–8)	5.98 (1.95)	5.74 (1.85)	0.24 *
CFI				0.99
TFI				0.98
RMSEA (90% CI)				0.03 (0.00–0.04)

Note. Standard deviations in parentheses. One-tailed tests. CFA = confirmatory factor analysis; CFI = comparative fit index; TFI = Tucker–Lewis Index; RMSEA = root mean square error of approximation. * p < 0.10. ** p < 0.05. *** p < 0.001.

report means and standard deviations of mothers’ responses to eight survey items on children’s behavior problems, by treatment status, in the pre-COVID sample. The mean scores of treatment children were higher than those of control children on all but two items: disobedience at school and disobedience at home. These results generally suggest that behavior problems were reported less frequently in the treatment group. The treatment–control differences were statistically significant at the level of 0.10 on three items: “The child complains no one loves him/her,” “The child is too fearful or anxious,” and “The child feels worthless or inferior.”

The mean of the sum score for these eight items was 0.24 higher for the treatment group (p < 0.10). The descriptive statistics suggest that behavior problems were reported less frequently by children in the treatment group than by their counterparts in the control group.

We tested a CFA model of children’s behavioral problems with two latent measures of anxiety and disobedience in the pre-COVID sample. The first five items focused on anxiety, and the remaining three focused on disobedience. This eight-item, two-factor CFA model had acceptable model-fit index values (comparative fit index, CFI = 0.99; Tucker–Lewis Index, TLI = 0.98; root mean square error of approximation, RMSEA = 0.03, 90% CI [0.00, 0.04]). All eight standardized factor loadings were statistically significant (p < 0.001), and six of them were greater than 0.60, higher than a conventional cutoff of 0.40 (Guadagnoli and Velicer 1988). The standardized factor loading on children’s anxiety for the item “The child has sudden changes in mood or feeling” was 0.59 (p < 0.001), very close to 0.60. However, the standardized factor loading on children’s anxiety was only 0.44 for the item “The child is too fearful or anxious” (p < 0.001). Overall, the CFA results suggest that these eight variables hold together well as a measure of children’s anxiety and disobedience.

3.3. CDA Effects on Adolescents’ Behavior Problems

The first panel of

Table 3. Regression results of CDA effects on children’s behavior problems.

CDA Effect	Pre-COVID Sample (n = 676)			Full Sample (N = 1712)
	Sum Score	Anxiety	Disobedience	Sum Score	Anxiety	Disobedience
Treatment group (yes)	0.24 (0.07) *	0.20 (0.04) **	0.03 (0.39)	0.04 (0.39)	0.05 (0.25)	−0.02 (0.38)
R2	0.07			0.07
CFI		0.97			0.96
TFI		0.95			0.94
RMSEA (90% CI)		0.02 (0.01–0.03)			0.02 (0.01–0.03)

Note. p values are reported in parentheses. One-tailed tests. CDA = Child Development Account; CFI = Comparative Fit Index; TFL = Tucker–Lewis Index; RMSEA = Root Mean Square Error Of Approximation; CI = Confidence Interval. * p < 0.10. ** p < 0.05.

reports the effects of CDAs on behavior problems of adolescents in the pre-COVID sample, showing regression coefficients and p-values. A positive coefficient indicates a lower frequency of behavior problems. With the inclusion of controls for Wave 1 demographic and socioeconomic characteristics, the mean sum score of child behavior problems was 0.24 higher for treatment children than for control children in the pre-COVID sample (p < 0.10), indicating that behavior problems were reported less frequently among the treatment children.

Regression analyses on two latent measures of anxiety and disobedience showed that CDAs primarily affected children’s anxiety. Among the pre-COVID sample, the mean value on the latent measure of anxiety for the treatment children was about 20% of a standard deviation higher—implying less frequent bouts of anxiety—than that for the control children (p < 0.05). Also, the mean value on the latent measure of disobedience was 3% of a standard deviation higher for treatment children than for control children, but this difference was not statistically significant. The second panel of Table 3 reports the effects of CDAs for the full sample at Wave 3. In those results, all three coefficients were statistically non-significant.

4. Discussion

This study builds upon previous SEED OK research on early childhood by examining the effects of CDAs on adolescent behavior problems. It used Wave 3 survey data from a longitudinal, randomized policy experiment with a statewide probability sample. The results for the pre-COVID sample suggest that asset building through CDAs reduces the frequency of children’s behavior problems. Overall, the study’s findings parallel those from previous SEED OK research on children’s social-emotional development among children aged 4 (Huang et al. 2014), but these new findings provide estimates of effects when SEED OK children were in adolescence. The positive effects of CDAs on behaviors in early childhood and in adolescence suggest that the effects may hold over time.

The analyses show that the sum score of behavior problems for treatment adolescents in the pre-COVID sample was about 24% of a standard deviation higher than that for counterparts in the control group, and this suggests that behavior problems were less frequent among treatment adolescents. The estimated size of the CDA’s effect on the social-emotional development of children aged 4 was about 8% to 16% of a standard deviation (Huang et al. 2014). The positive effects found in both Wave 2 and Wave 3 suggest that the influences of this low-touch CDA policy on children and families were sustained. It is notable that these CDA impacts occurred and might persist even though the funds in the treatment children’s CDAs have not yet been spent on postsecondary education. In standardizing the analysis results on the sum score, we found that the regression coefficient of treatment status, 0.07 (p < 0.10), is comparable with those of children’s gender (0.06, p < 0.10) and household income-to-needs ratio (0.11, p < 0.05). The effect size of CDAs is noteworthy because several control variables (mother’s education, employment, homeownership, TANF participation, and SNAP participation) in these analyses were not found to have statistically significant associations with the behavioral problems of children at age 13. One would expect these socioeconomic characteristics to matter for child behavior.

Different measures of adolescents’ behavior problems (sum score and latent measures of anxiety and disobedience) may be related to different estimates of CDA impacts for children. For example, the analyses on anxiety and disobedience suggest that the statistically significant result on the sum score was mainly due to the CDA effects on adolescents’ internalized behavior problems. Results from the CFA measures of anxiety and disobedience suggest that the effects of CDAs may differ by types of adolescents’ behavior problems. The effects on children’s anxiety were greater than those on disobedience. The mean score of the latent measure of children’s anxiety in the treatment group was about one-fifth (20%) of a standard deviation higher than that in the control group.

Previous research (Huang et al. 2021; Sherraden et al. 2018) from the SEED OK experiment has firmly established the financial and psychosocial benefits of CDAs, linking them to increased family savings (Huang et al. 2021), reduced maternal depression (Huang et al. 2014), and more positive parenting styles (Huang et al. 2014). This study builds directly on that foundation by asking if these stabilizing family effects translate into fewer behavior problems for children as they enter adolescence.

Our analysis contributes a critical new piece to this literature: evidence that the influence of CDAs on psychosocial well-being persists into adolescence, manifesting as a measurable reduction in behavior problems. These effects—positive impacts on behaviors of children aged 12 and 13—should not be considered in isolation but rather as part of a larger set of findings on asset building’s long-term effects on children’s health, well-being, and development. The body of findings from SEED OK is consistent with the hypotheses of asset-effects theory (Sherraden 1991): that lifelong asset building can have long-term psychosocial and health benefits for families and children.

These findings have important implications for future research and policy promoting asset building for children’s health and well-being. Such an understanding may contribute to the research, design, and implementation of future CDAs in the United States and across the globe. For example, it should be noted that the positive effects of CDAs on adolescent behavior problems are generated from a specific form of universal and progressive CDA design; its effects may vary with differences in design. The findings on specific CDA effects by types of behavior problems could be used to improve policy design. If CDAs have stronger impacts on children’s anxiety than on disobedience, some financial incentives in CDAs could be tied to specific children’s behaviors. Integrating mental health services for anxiety and stress into asset-building policies may maximize CDA impacts on child development.

As asset-building policy gains traction, seven states have now implemented universal CDAs based on the SEED OK framework. The findings from this study are particularly salient for these new programs, as we provide evidence that the CDA model can serve as a preventative mental health strategy. By reducing behavior problems, particularly anxiety, these statewide policies may generate significant societal returns by fostering a generation of more resilient and behaviorally healthy adolescents. Further research is needed to systematically examine the potential of universal CDA policy as a comprehensive system for adolescent development, including research on other aspects of behavioral health not examined in the current study.

Although SEED OK is a randomized experiment with a state probability sample, this study has some limitations. First, the Wave 3 data collection was disrupted by the COVID pandemic, which may have affected results for the full sample. Those results indicated that CDAs have statistically non-significant impacts on adolescent behavior. Because of this potential for disruption, our analyses focused on the results for the pre-COVID sample. A similar strategy has been applied in other research (Troller-Renfree et al. 2022); however, the extent of the pandemic’s effects remains unclear. It likely disrupted people’s well-being and outlook in profound, if unmeasured, ways. The long-term CDA impacts on the full sample in SEED OK should be further evaluated in the future with better data. A dedicated future analysis is to specifically investigate the post-COVID data, aiming to disentangle the intervention’s effects from the pandemic’s profound influence on both participant well-being and data quality. Nonetheless, the consistently positive effects of CDAs on behavior problems, effects observed at Wave 2 and in the pre-COVID sample, offer support for our confidence in the asset-effect hypotheses.

Second, the CFA results suggest that different behavior outcome measures lead to different estimates of CDA effects, highlighting the importance of understanding the specific mechanisms of asset effects for children. Differences in estimates on anxiety and disobedience indicate that CDA effects may vary by behavior problem. With the current research design, this study offers only a reduced-form assessment of the CDA’s direct effects on adolescent behavior.

Third, approximately 37% of the SEED OK Wave 1 participants were lost to attrition by the time of the Wave 3 survey. Descriptive analyses indicated that the treatment and control groups, when reweighted to represent the original population after attrition, are well balanced and comparable in terms of demographic and socioeconomic characteristics. This is true for the pre-COVID and full samples. However, the sample attrition may create challenges for the external validity of the study.

Despite its limitations, this study is the first to provide longitudinal follow-up evidence concerning the effects of CDAs on children’s behavior problems in the time since the children were aged 4. The number of children with a CDA has grown rapidly. As of 2021, over 5 million children in the United States held assets in a CDA (Zou and Sherraden 2022). By 2025, this figure had increased to approximately 8 million. The 2021 estimate for children with CDAs globally was over 15 million (Zou and Sherraden 2022). As more children are included in CDA policies, research evidence on how CDAs affect children’s well-being and development is crucial for policy implementation, evaluation, and development. Our research on SEED OK showed that CDAs have positive effects on both financial and nonfinancial outcomes (Huang et al. 2021). The findings of this study provide additional evidence to support and promote asset building for all children.

5. Conclusions

In summary, our study has yielded promising results, revealing the positive effects of CDAs on adolescent behavioral health. These findings highlight the potential of CDAs as a valuable tool not only for enhancing financial well-being but also for promoting positive behavior outcomes among adolescents. The findings underscore the broader significance of asset-building policies in nurturing resilient and healthy individuals, with implications for enhancing the well-being of our youth and communities. Moving forward, our study encourages continued exploration and refinement of such policies to empower future generations.