TMU
Owner
Tarbiat Modares University (TMU) is the owner of HEHP.
0.8
Cite Score
Volume 13, Issue 3 (2025)
Health Educ Health Promot 2025, 13(3): 429-436 |
Back to browse issues page
History
Received: 2025/04/21 | Accepted: 2025/07/1 | Published: 2025/07/16
Received: 2025/04/21 | Accepted: 2025/07/1 | Published: 2025/07/16
How to cite this article
Effect of Family Structure on the Non-Cognitive Ability of Rural Adolescents. Health Educ Health Promot 2025; 13 (3) :429-436
URL: http://hehp.modares.ac.ir/article-4-80457-en.html
URL: http://hehp.modares.ac.ir/article-4-80457-en.html
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Rights and permissions
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Authors
Keywords:
| Abstract (HTML) (797 Views)
Full-Text: (149 Views)
Introduction
Widespread and complex transformations are occurring in Chinese family structures. According to statistics, the divorce rate has increased by 36% over the past twenty years (from 1.91% in 2000 to 2.6% in 2022), which has driven the number of single-parent families in China above 24 million. While rural populations are experiencing the unwanted consequences of urbanization—specifically, the displacement of sixty million children into their grandparents’ homes [1]—this practice is increasingly formalized as skip-generation parentification. Such configurations run counter to deteriorating scores on mental health measures, with 14% and 40% of adolescents displaying signs of depression and severe depression, respectively [2]. Among the various nuclear family configurations—such as remarried families, grandparent-assisted families, and lone-parent families—the pathways through which these configurations reshape adolescents’ resilience, cooperative capacities, and emotional well-being are crucial to understand. This understanding is essential for cultivating China’s “population quality dividend,” which is central to the country’s developmental calculus.
These considerations manifested within policy realms: The “Double Reduction” initiative realigns pedagogic burdens toward the home, and the Family Education Promotion Law institutes directives regarding adolescents’ psychological welfare. Crucially, the 2023 Action Plan for Strengthening Student Mental Health, released by seventeen ministries, recognizes the aforementioned changes in family structure as potential mental health risk factors. However, these policies remain hindered by conceptual conservatism. They rest upon fixed notions of what constitutes a family unit and consequently overlook structural multiplicity, particularly regarding single-parent dynamics, skip-generation dependency, and nuclear or extended arrangements. This failure creates significant gaps in evidence. For instance, the psychosocial channels that mediate these structural impacts remain unclear. Additionally, the effects of socioeconomic gradients and the rural-urban divide on outcomes are not well understood. Without this information, governance may risk implementing fragmented solutions that do not address the systemic nature of the issues.
In parallel, academic scholarship also reflects similar deficiencies. Prevalent inquiries anchored in family capital theory reduce its influence to economic pipelines and repositories [3-5]. In doing so, they occlude the self-will of structure. A study tracking a sample of Chinese children growing up in rural extended family settings reveals that such relationships can partly compensate for deficiencies in cooperative skills, but only when elders receive adequate education [6-8]. Consequently, two analytical limitations become apparent. First, measurement systems tend to rely on simplistic “intact/broken” classifications, failing to account for the differences between maternal and paternal absences and the intricate division of roles among siblings. Second, explanatory frameworks cling to overly simplistic linear “structure-resource-outcome” sequences, disregarding the emotional factors at play, such as adolescents’ emotional reactions to conflict resolution, reciprocal learning processes, competition within sibling bonds, and the influence of role models in brother-sister interactions, as well as the inherent competitive interdependence in sibling pairs. Although official bodies focus on financial measures, delving into the emotional dimensions of family dynamics might offer more significant insights into these complexities..
Family structure, as a central context in adolescent socialization, has long been studied across education, sociology, and psychology for its impact on non-cognitive skills. Traditional research grounded in family capital theory emphasizes economic capital (income, assets) and cultural capital (parental education, cultural practices) as primary drivers of youth development [9-12]. Yet, this framework struggles to explain paradoxical findings: Adolescents from economically disadvantaged rural extended families often exhibit greater emotional stability than those from urban nuclear households [13], while adolescents in high-income single-parent families show lower cooperative tendencies than those in low-income intact families [14-16]. Such evidence suggests structural influences operate independently of material resources.
Family systems theory offers fresh insights by highlighting the structural effects of role allocation, intergenerational dynamics, and emotional support networks. For example, multi-generational households may strengthen cooperative skills through kinship bonds, but intergenerational parenting conflicts could hinder autonomy [17, 18]. Social capital theory further posits that family structure modulates non-cognitive skill development through network density (e.g., sibling count) and relational quality (e.g., parent-child communication frequency). Only children might develop weaker resilience due to the absence of sibling rivalry, although intensive parental interaction could offset this deficit [19-21].
Collectively, these theories outline a three-tiered “structure-interaction-competence” pathway, although empirical validation of specific mechanisms remains limited. Recent advancements in life course theory [22] have illuminated the dynamic effects of family structural changes (such as divorce or the birth of a second child) on non-cognitive skill development. Empirical studies reveal a marked decline in adolescents’ emotional stability during the initial phase of parental divorce [23-25], although long-term adaptation may occur through new support systems formed during family reorganization [26, 27]. While these findings collectively outline a three-tiered “structure-interaction-ability” framework, critical gaps persist in empirically validating its underlying mechanisms.
Prevailing studies often employ simplistic “intact versus single-parent” classifications, overlooking structural heterogeneity. The absence of parents can manifest in various forms: fatherless households are associated with reduced perseverance [28], while families lacking maternal presence exhibit a stronger correlation with emotional instability [29]. Similarly, studies on skip-generation caregiving frequently neglect the educational disparities among grandparents [30, 31]. Although grandparents with higher education levels may mitigate emotional detachment through intellectual engagement, those with lower educational backgrounds might inadvertently foster dependency by overprotecting their grandchildren.
The existing literature overstates the allocative mechanism while underestimating emotional interaction dynamics. The economics of extended families could be offset by intergenerational tensions, and the frequency of dialogue within nuclear families may enhance logic but hinder decision-making independence due to frequent intervention [32, 33]. There is conflicting evidence regarding role modeling: some studies indicate that responsibility spillover occurs through the older sibling effect [34], while others find less cooperation as siblings compete for resources [35].
Three critical limitations persist in existing research. First, binary static classifications (“intact vs. broken”) fail to differentiate between the effects of maternal and paternal absence or the educational levels of grandparents in skip-generation care. Second, mechanistic explanations overly rely on economic capital, neglecting psychosocial pathways, such as emotional reciprocity and intergenerational modeling. Third, policy designs treat families as homogeneous units, lacking structural nuance. These gaps hinder explanations for paradoxical outcomes (e.g., rural extended families’ emotional stability advantages) and impede evidence-based interventions.
This study addressed these gaps through theoretical synthesis and methodological innovation. First, we developed a multidimensional family structure typology using longitudinal data from the China Education Panel Survey (CEPS) to distinguish between guardianship types (dual-parent and single-parent). Second, moving beyond resource determinism, we examined mechanisms in non-traditional structures, such as multi-child and skip-generation households. Third, we advocate shifting the policy focus from economic compensation to capacity-building interventions, such as maternal custody support and grandparent caregiver training. By integrating CEPS multi-source data, refining parameter operationalization, and employing multilevel modeling, this framework systematically evaluated the heterogeneous impacts of family structure and its contextual boundaries.
Instrument and Methods
Study design, data source, and sample
This study utilized longitudinal data from the CEPS, incorporating both the 2013-2014 baseline and the 2014-2015 follow-up waves. To ensure temporal consistency, we performed exact matching of participants across survey waves using Stata’s -reclink- command with unique student identifiers (IDS), retaining only rural students present in both waves (n=5,111 after quality control).
The data integration process leveraged Stata’s -merge- and -joinby- commands to combine four data sources, including student self-reports (ID-stu), parent questionnaires (ID-par), homeroom teacher evaluations (ID-tch), and school administrative records (ID-sch). We implemented Stata’s -mi- commands to evaluate missing data patterns, with Little’s MCAR test (-mcartest- user-written command) confirming that the missing data patterns were completely at random (χ²=18.34, p=0.12). Cases exceeding 30% missingness on core parameters (family structure and non-cognitive skills) were excluded using Stata’s -drop- if- conditional statements. All procedures performed in the original study were in accordance with the ethical standards of the institutional and national research committees and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.
For the assessment of sample representativeness, we employed Stata’s -teffects- suite to calculate inverse probability weights, with all standardized mean differences (<0.10) verified using the stddiff package. The -svyset- command established appropriate survey weights for subsequent analyses. This comprehensive Stata-based workflow ensured rigorous data management and preparation for our analytical models.
Non-cognitive skills were operationalized using the Big Five personality model, a widely validated framework in research on Chinese adolescents [36]. The five dimensions of conscientiousness, extraversion, agreeableness, openness, and emotional stability were measured. First, principal component analysis (PCA) was applied to the questionnaire items within each dimension. All constructs demonstrated adequate sampling adequacy (KMO>0.6) and significant Bartlett’s tests (p<0.05). Second, PCA-derived factor scores were standardized to a 0-1 scale. Third, dimension scores were averaged, with higher values indicating stronger non-cognitive skills (Appendix 1 and Table 1).
Table 1. Descriptive statistics (n=5111)
![src=./files/hehp/images/HTML_Publish/80457/1.png]()
Family structure, a core explanatory parameter, was assessed using the 2013-2014 CEPS item that identifies which immediate family members do not currently reside with the child. Following established protocols [37, 38], we implemented two operationalizations, including a binary parameter (0=dual-parent residency; 1=other, including single-parent or non-parent residency) and a four-category parameter: 1=dual-parent, 2=mother-only, 3=father-only, 4=non-parent residency (e.g., skip-generation care). Dual-parent households served as the reference group in all analyses.
Covariates were categorized as follows: student: age, gender, singleton status; parental: maternal/paternal education levels, parental relationship status; household: economic status, educational expectations, log-transformed extracurricular expenditure; School: log-transformed per-student funding, college admission rate. Regions were included to control for geographic heterogeneity.
The analysis employed a linear regression model with robust standard errors, controlling for demographic and socioeconomic covariates. All data processing and statistical analyses were conducted using Stata/MP 17 (StataCorp LLC, College Station, TX).
Findings
The sample comprised 5,111 adolescents, with the majority (72.59%, n=3,710) residing in dual-parent households. Non-parent households constituted 14.03% (n=717), while mother-only and father-only arrangements represented 9.61% (n=491) and 3.78% (n=193), respectively.
Adolescents in non-parent-resident households exhibited significantly lower non-cognitive skills compared to their dual-parent counterparts (β=-0.025, SE=0.006, t=-4.36, p<0.01), representing a 2.5% reduction in standardized non-cognitive scores. Mother-only households showed a statistically significant negative association (β=-0.020, SE=0.006, t=-3.12, p=0.002), although the magnitude was marginally smaller than that of non-parent households.
The effect size for father-only households (β=-0.022, SE=0.010, t=-2.18, p=0.029) approached statistical significance but demonstrated wider confidence intervals, likely due to a smaller sample size (n=193, 3.78% of the total).
The regression model demonstrated satisfactory explanatory power (adjusted R²=0.104; R²=0.111), indicating that approximately 10.4% of the variance in non-cognitive skills was attributable to the specified predictors. The root mean square error (RMSE=0.050) further confirmed model adequacy.
When considering gendered caregiving architecture, cultural scripts codified mothers as chief nurturing authorities (architects of emotional scaffolding and behavioral regulation), while fathers assumed fiscal provisioning roles. Rural paternal labor patterns, whether agrarian or non-agrarian, constrained childcare investments in terms of time due to protracted labor practices. The empirical consensus corroborated this directionality, with adolescents in lone-parent and non-parent configurations demonstrating statistically lower competencies compared to those in dual-parent family formations (Table 2).
Table 2. Regression findings
![src=./files/hehp/images/HTML_Publish/80457/2.png]()
Family structure impacted non-cognitive skills differently for only children and those with siblings. For only children, non-cognitive outcomes remained stable with at least one parent present, but full parental absence led to a sharp decline (β=-0.055, p<0.01), highlighting their vulnerability. In contrast, non-only children faced deficits even with partial parental absence, with mother-only (β=-0.025, p<0.01) and father-only (β=-0.024, p<0.05) households showing negative effects. Surprisingly, full parental absence had a lesser impact on non-only children (β=-0.018, p<0.01), likely due to sibling support (Table 3).
Table 3. Heterogeneity analysis (if the only child)
![src=./files/hehp/images/HTML_Publish/80457/3.png]()
In binary models, parental absence had a stronger negative impact on non-cognitive abilities when grandparents co-reside (β=-0.020, p<0.001) compared to when they do not (β=-0.016, p<0.01).
Maternal care without grandparents correlated with modest skill deficits (β=-0.013, p<0.1), while co-residence further reduced the effect size and significance (β=-0.012). In father-only households, grandparental support was even more critical; paternal care alone resulted in non-significant deficits (β=-0.016; p<0.01), but cohabitation with grandparents almost nullified negative effects (β=-0.002).
By contrast, non-parent households with grandparental care showed the most pronounced deficits (β=-0.026, p<0.05 vs. β=-0.019, p<0.01), highlighting grandparents’ limited ability to fully substitute for parental roles (Table 4).
Table 4. Heterogeneity analysis (if living with grandparents)
![src=./files/hehp/images/HTML_Publish/80457/4.png]()
Discussion
This study explored the diverse impacts of family structure on rural Chinese adolescents’ non-cognitive skills using longitudinal data from the CEPS. We systematically revealed the heterogeneous impacts of family structures on rural adolescents’ non-cognitive skills and the mechanisms defining their developmental boundaries. Beyond reaffirming the central role of family structure, it advances theoretical understanding by deconstructing the pathways of non-cognitive ability formation.
Consistent with Cui & Xu’s study [39], we found that adolescents in intact dual-parent families performed best in non-cognitive domains, whereas parental absence undermined emotional stability and cooperation more severely. Crucially, intra-structural heterogeneity emerges: family structure operates through the interaction of guardianship roles, emotional reciprocity, and cultural capital. Maternal absence inflicts greater setbacks than paternal absence, reflecting gendered caregiving dynamics rooted in rural “motherhood scripts” [40]. Fathers prioritize economic provision and behavioral control, while grandparents rely on outdated educational norms emphasizing obedience, often creating emotional rifts in non-cognitive development, particularly in emotional stability and cooperation [40-43]. These findings challenge the economic determinism of family capital theory, suggesting that emotional capital may yield greater marginal returns than material resources.
The impact of family structure must be contextualized within life-course dynamics. The duration, cause (e.g., divorce, labor migration, death), and timing of parental absence jointly shape non-cognitive trajectories. Labor migration-induced absence initially triggers acute separation anxiety (<6 months), but after 24 months, some adolescents adapt through alternative support networks such as peer bonds and teacher mentorship [44, 45]. This nonlinearity introduces the concept of resilience thresholds (points beyond which external support mitigates structural deficits), challenging linear assumptions of “absence-duration=harm” [46]. Although the divergent roles of mothers and fathers are clarified, their interplay across dynamic processes warrants longitudinal exploration.
While research [47] observes academic benefits from grandparent co-residence, our results indicated that, in rural contexts, elders’ lower educational attainment and traditional norms intensified non-cognitive deficits in the absence of parents, aligning with Yu’s study [48]. This difference may stem from China’s unique national conditions. Physical and cognitive limitations could also hinder grandparental supervision of adolescents’ out-of-school activities, such as Internet use and peer interaction, thus impeding self-regulation. Effective intergenerational support requires strengthening caregiver competencies rather than assuming that co-residence itself ensures quality caregiving.
Finally, sibling interdependencies expose notable subcohort divergences. Consistent with broader evidence of sibling effects, a study of rural children in northwest China demonstrates two countervailing mechanisms—“interaction engagement” versus “resource dilution”—at work as sibling count changes [49]. When only one sibling is present, enhanced communication and emotional support (“interaction engagement”) confer measurable advantages in psychological well-being and academic performance compared to only children. However, once families have two or more siblings, competition for parental attention and material resources (“resource dilution”) leads to significant declines in both domains.
Integrating these dynamics into our findings, we observe that in multi-child households under migration pressures, elder-sibling caregiving can indeed foster early cooperative skills and emotional maturity (the “interaction engagement” effect), but this benefit is counterbalanced by increased role strain and scarce resources when sibling numbers exceed one. This nuanced view cautions against assuming uniformly protective effects of sibling social capital, underscoring that both the quality of interaction and the quantity of siblings jointly shape non-cognitive outcomes.
The study’s limitations offer important directions for future research. First, the CEPS data lack granularity in measuring family structure heterogeneity, failing to distinguish the causes of parental absence or capture dynamic changes. Future studies could integrate in-depth interviews to develop multidimensional structu
Widespread and complex transformations are occurring in Chinese family structures. According to statistics, the divorce rate has increased by 36% over the past twenty years (from 1.91% in 2000 to 2.6% in 2022), which has driven the number of single-parent families in China above 24 million. While rural populations are experiencing the unwanted consequences of urbanization—specifically, the displacement of sixty million children into their grandparents’ homes [1]—this practice is increasingly formalized as skip-generation parentification. Such configurations run counter to deteriorating scores on mental health measures, with 14% and 40% of adolescents displaying signs of depression and severe depression, respectively [2]. Among the various nuclear family configurations—such as remarried families, grandparent-assisted families, and lone-parent families—the pathways through which these configurations reshape adolescents’ resilience, cooperative capacities, and emotional well-being are crucial to understand. This understanding is essential for cultivating China’s “population quality dividend,” which is central to the country’s developmental calculus.
These considerations manifested within policy realms: The “Double Reduction” initiative realigns pedagogic burdens toward the home, and the Family Education Promotion Law institutes directives regarding adolescents’ psychological welfare. Crucially, the 2023 Action Plan for Strengthening Student Mental Health, released by seventeen ministries, recognizes the aforementioned changes in family structure as potential mental health risk factors. However, these policies remain hindered by conceptual conservatism. They rest upon fixed notions of what constitutes a family unit and consequently overlook structural multiplicity, particularly regarding single-parent dynamics, skip-generation dependency, and nuclear or extended arrangements. This failure creates significant gaps in evidence. For instance, the psychosocial channels that mediate these structural impacts remain unclear. Additionally, the effects of socioeconomic gradients and the rural-urban divide on outcomes are not well understood. Without this information, governance may risk implementing fragmented solutions that do not address the systemic nature of the issues.
In parallel, academic scholarship also reflects similar deficiencies. Prevalent inquiries anchored in family capital theory reduce its influence to economic pipelines and repositories [3-5]. In doing so, they occlude the self-will of structure. A study tracking a sample of Chinese children growing up in rural extended family settings reveals that such relationships can partly compensate for deficiencies in cooperative skills, but only when elders receive adequate education [6-8]. Consequently, two analytical limitations become apparent. First, measurement systems tend to rely on simplistic “intact/broken” classifications, failing to account for the differences between maternal and paternal absences and the intricate division of roles among siblings. Second, explanatory frameworks cling to overly simplistic linear “structure-resource-outcome” sequences, disregarding the emotional factors at play, such as adolescents’ emotional reactions to conflict resolution, reciprocal learning processes, competition within sibling bonds, and the influence of role models in brother-sister interactions, as well as the inherent competitive interdependence in sibling pairs. Although official bodies focus on financial measures, delving into the emotional dimensions of family dynamics might offer more significant insights into these complexities..
Family structure, as a central context in adolescent socialization, has long been studied across education, sociology, and psychology for its impact on non-cognitive skills. Traditional research grounded in family capital theory emphasizes economic capital (income, assets) and cultural capital (parental education, cultural practices) as primary drivers of youth development [9-12]. Yet, this framework struggles to explain paradoxical findings: Adolescents from economically disadvantaged rural extended families often exhibit greater emotional stability than those from urban nuclear households [13], while adolescents in high-income single-parent families show lower cooperative tendencies than those in low-income intact families [14-16]. Such evidence suggests structural influences operate independently of material resources.
Family systems theory offers fresh insights by highlighting the structural effects of role allocation, intergenerational dynamics, and emotional support networks. For example, multi-generational households may strengthen cooperative skills through kinship bonds, but intergenerational parenting conflicts could hinder autonomy [17, 18]. Social capital theory further posits that family structure modulates non-cognitive skill development through network density (e.g., sibling count) and relational quality (e.g., parent-child communication frequency). Only children might develop weaker resilience due to the absence of sibling rivalry, although intensive parental interaction could offset this deficit [19-21].
Collectively, these theories outline a three-tiered “structure-interaction-competence” pathway, although empirical validation of specific mechanisms remains limited. Recent advancements in life course theory [22] have illuminated the dynamic effects of family structural changes (such as divorce or the birth of a second child) on non-cognitive skill development. Empirical studies reveal a marked decline in adolescents’ emotional stability during the initial phase of parental divorce [23-25], although long-term adaptation may occur through new support systems formed during family reorganization [26, 27]. While these findings collectively outline a three-tiered “structure-interaction-ability” framework, critical gaps persist in empirically validating its underlying mechanisms.
Prevailing studies often employ simplistic “intact versus single-parent” classifications, overlooking structural heterogeneity. The absence of parents can manifest in various forms: fatherless households are associated with reduced perseverance [28], while families lacking maternal presence exhibit a stronger correlation with emotional instability [29]. Similarly, studies on skip-generation caregiving frequently neglect the educational disparities among grandparents [30, 31]. Although grandparents with higher education levels may mitigate emotional detachment through intellectual engagement, those with lower educational backgrounds might inadvertently foster dependency by overprotecting their grandchildren.
The existing literature overstates the allocative mechanism while underestimating emotional interaction dynamics. The economics of extended families could be offset by intergenerational tensions, and the frequency of dialogue within nuclear families may enhance logic but hinder decision-making independence due to frequent intervention [32, 33]. There is conflicting evidence regarding role modeling: some studies indicate that responsibility spillover occurs through the older sibling effect [34], while others find less cooperation as siblings compete for resources [35].
Three critical limitations persist in existing research. First, binary static classifications (“intact vs. broken”) fail to differentiate between the effects of maternal and paternal absence or the educational levels of grandparents in skip-generation care. Second, mechanistic explanations overly rely on economic capital, neglecting psychosocial pathways, such as emotional reciprocity and intergenerational modeling. Third, policy designs treat families as homogeneous units, lacking structural nuance. These gaps hinder explanations for paradoxical outcomes (e.g., rural extended families’ emotional stability advantages) and impede evidence-based interventions.
This study addressed these gaps through theoretical synthesis and methodological innovation. First, we developed a multidimensional family structure typology using longitudinal data from the China Education Panel Survey (CEPS) to distinguish between guardianship types (dual-parent and single-parent). Second, moving beyond resource determinism, we examined mechanisms in non-traditional structures, such as multi-child and skip-generation households. Third, we advocate shifting the policy focus from economic compensation to capacity-building interventions, such as maternal custody support and grandparent caregiver training. By integrating CEPS multi-source data, refining parameter operationalization, and employing multilevel modeling, this framework systematically evaluated the heterogeneous impacts of family structure and its contextual boundaries.
Instrument and Methods
Study design, data source, and sample
This study utilized longitudinal data from the CEPS, incorporating both the 2013-2014 baseline and the 2014-2015 follow-up waves. To ensure temporal consistency, we performed exact matching of participants across survey waves using Stata’s -reclink- command with unique student identifiers (IDS), retaining only rural students present in both waves (n=5,111 after quality control).
The data integration process leveraged Stata’s -merge- and -joinby- commands to combine four data sources, including student self-reports (ID-stu), parent questionnaires (ID-par), homeroom teacher evaluations (ID-tch), and school administrative records (ID-sch). We implemented Stata’s -mi- commands to evaluate missing data patterns, with Little’s MCAR test (-mcartest- user-written command) confirming that the missing data patterns were completely at random (χ²=18.34, p=0.12). Cases exceeding 30% missingness on core parameters (family structure and non-cognitive skills) were excluded using Stata’s -drop- if- conditional statements. All procedures performed in the original study were in accordance with the ethical standards of the institutional and national research committees and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.
For the assessment of sample representativeness, we employed Stata’s -teffects- suite to calculate inverse probability weights, with all standardized mean differences (<0.10) verified using the stddiff package. The -svyset- command established appropriate survey weights for subsequent analyses. This comprehensive Stata-based workflow ensured rigorous data management and preparation for our analytical models.
Non-cognitive skills were operationalized using the Big Five personality model, a widely validated framework in research on Chinese adolescents [36]. The five dimensions of conscientiousness, extraversion, agreeableness, openness, and emotional stability were measured. First, principal component analysis (PCA) was applied to the questionnaire items within each dimension. All constructs demonstrated adequate sampling adequacy (KMO>0.6) and significant Bartlett’s tests (p<0.05). Second, PCA-derived factor scores were standardized to a 0-1 scale. Third, dimension scores were averaged, with higher values indicating stronger non-cognitive skills (Appendix 1 and Table 1).
Table 1. Descriptive statistics (n=5111)
Family structure, a core explanatory parameter, was assessed using the 2013-2014 CEPS item that identifies which immediate family members do not currently reside with the child. Following established protocols [37, 38], we implemented two operationalizations, including a binary parameter (0=dual-parent residency; 1=other, including single-parent or non-parent residency) and a four-category parameter: 1=dual-parent, 2=mother-only, 3=father-only, 4=non-parent residency (e.g., skip-generation care). Dual-parent households served as the reference group in all analyses.
Covariates were categorized as follows: student: age, gender, singleton status; parental: maternal/paternal education levels, parental relationship status; household: economic status, educational expectations, log-transformed extracurricular expenditure; School: log-transformed per-student funding, college admission rate. Regions were included to control for geographic heterogeneity.
The analysis employed a linear regression model with robust standard errors, controlling for demographic and socioeconomic covariates. All data processing and statistical analyses were conducted using Stata/MP 17 (StataCorp LLC, College Station, TX).
Findings
The sample comprised 5,111 adolescents, with the majority (72.59%, n=3,710) residing in dual-parent households. Non-parent households constituted 14.03% (n=717), while mother-only and father-only arrangements represented 9.61% (n=491) and 3.78% (n=193), respectively.
Adolescents in non-parent-resident households exhibited significantly lower non-cognitive skills compared to their dual-parent counterparts (β=-0.025, SE=0.006, t=-4.36, p<0.01), representing a 2.5% reduction in standardized non-cognitive scores. Mother-only households showed a statistically significant negative association (β=-0.020, SE=0.006, t=-3.12, p=0.002), although the magnitude was marginally smaller than that of non-parent households.
The effect size for father-only households (β=-0.022, SE=0.010, t=-2.18, p=0.029) approached statistical significance but demonstrated wider confidence intervals, likely due to a smaller sample size (n=193, 3.78% of the total).
The regression model demonstrated satisfactory explanatory power (adjusted R²=0.104; R²=0.111), indicating that approximately 10.4% of the variance in non-cognitive skills was attributable to the specified predictors. The root mean square error (RMSE=0.050) further confirmed model adequacy.
When considering gendered caregiving architecture, cultural scripts codified mothers as chief nurturing authorities (architects of emotional scaffolding and behavioral regulation), while fathers assumed fiscal provisioning roles. Rural paternal labor patterns, whether agrarian or non-agrarian, constrained childcare investments in terms of time due to protracted labor practices. The empirical consensus corroborated this directionality, with adolescents in lone-parent and non-parent configurations demonstrating statistically lower competencies compared to those in dual-parent family formations (Table 2).
Table 2. Regression findings
Family structure impacted non-cognitive skills differently for only children and those with siblings. For only children, non-cognitive outcomes remained stable with at least one parent present, but full parental absence led to a sharp decline (β=-0.055, p<0.01), highlighting their vulnerability. In contrast, non-only children faced deficits even with partial parental absence, with mother-only (β=-0.025, p<0.01) and father-only (β=-0.024, p<0.05) households showing negative effects. Surprisingly, full parental absence had a lesser impact on non-only children (β=-0.018, p<0.01), likely due to sibling support (Table 3).
Table 3. Heterogeneity analysis (if the only child)
In binary models, parental absence had a stronger negative impact on non-cognitive abilities when grandparents co-reside (β=-0.020, p<0.001) compared to when they do not (β=-0.016, p<0.01).
Maternal care without grandparents correlated with modest skill deficits (β=-0.013, p<0.1), while co-residence further reduced the effect size and significance (β=-0.012). In father-only households, grandparental support was even more critical; paternal care alone resulted in non-significant deficits (β=-0.016; p<0.01), but cohabitation with grandparents almost nullified negative effects (β=-0.002).
By contrast, non-parent households with grandparental care showed the most pronounced deficits (β=-0.026, p<0.05 vs. β=-0.019, p<0.01), highlighting grandparents’ limited ability to fully substitute for parental roles (Table 4).
Table 4. Heterogeneity analysis (if living with grandparents)
Discussion
This study explored the diverse impacts of family structure on rural Chinese adolescents’ non-cognitive skills using longitudinal data from the CEPS. We systematically revealed the heterogeneous impacts of family structures on rural adolescents’ non-cognitive skills and the mechanisms defining their developmental boundaries. Beyond reaffirming the central role of family structure, it advances theoretical understanding by deconstructing the pathways of non-cognitive ability formation.
Consistent with Cui & Xu’s study [39], we found that adolescents in intact dual-parent families performed best in non-cognitive domains, whereas parental absence undermined emotional stability and cooperation more severely. Crucially, intra-structural heterogeneity emerges: family structure operates through the interaction of guardianship roles, emotional reciprocity, and cultural capital. Maternal absence inflicts greater setbacks than paternal absence, reflecting gendered caregiving dynamics rooted in rural “motherhood scripts” [40]. Fathers prioritize economic provision and behavioral control, while grandparents rely on outdated educational norms emphasizing obedience, often creating emotional rifts in non-cognitive development, particularly in emotional stability and cooperation [40-43]. These findings challenge the economic determinism of family capital theory, suggesting that emotional capital may yield greater marginal returns than material resources.
The impact of family structure must be contextualized within life-course dynamics. The duration, cause (e.g., divorce, labor migration, death), and timing of parental absence jointly shape non-cognitive trajectories. Labor migration-induced absence initially triggers acute separation anxiety (<6 months), but after 24 months, some adolescents adapt through alternative support networks such as peer bonds and teacher mentorship [44, 45]. This nonlinearity introduces the concept of resilience thresholds (points beyond which external support mitigates structural deficits), challenging linear assumptions of “absence-duration=harm” [46]. Although the divergent roles of mothers and fathers are clarified, their interplay across dynamic processes warrants longitudinal exploration.
While research [47] observes academic benefits from grandparent co-residence, our results indicated that, in rural contexts, elders’ lower educational attainment and traditional norms intensified non-cognitive deficits in the absence of parents, aligning with Yu’s study [48]. This difference may stem from China’s unique national conditions. Physical and cognitive limitations could also hinder grandparental supervision of adolescents’ out-of-school activities, such as Internet use and peer interaction, thus impeding self-regulation. Effective intergenerational support requires strengthening caregiver competencies rather than assuming that co-residence itself ensures quality caregiving.
Finally, sibling interdependencies expose notable subcohort divergences. Consistent with broader evidence of sibling effects, a study of rural children in northwest China demonstrates two countervailing mechanisms—“interaction engagement” versus “resource dilution”—at work as sibling count changes [49]. When only one sibling is present, enhanced communication and emotional support (“interaction engagement”) confer measurable advantages in psychological well-being and academic performance compared to only children. However, once families have two or more siblings, competition for parental attention and material resources (“resource dilution”) leads to significant declines in both domains.
Integrating these dynamics into our findings, we observe that in multi-child households under migration pressures, elder-sibling caregiving can indeed foster early cooperative skills and emotional maturity (the “interaction engagement” effect), but this benefit is counterbalanced by increased role strain and scarce resources when sibling numbers exceed one. This nuanced view cautions against assuming uniformly protective effects of sibling social capital, underscoring that both the quality of interaction and the quantity of siblings jointly shape non-cognitive outcomes.
The study’s limitations offer important directions for future research. First, the CEPS data lack granularity in measuring family structure heterogeneity, failing to distinguish the causes of parental absence or capture dynamic changes. Future studies could integrate in-depth interviews to develop multidimensional structu