Mechanisms linking family socioeconomic status to children’s depressive symptoms: a longitudinal moderated mediation analysis

Abstract

Background

Family socioeconomic status (SES) is a well-established determinant of children’s mental health, yet the mechanisms linking SES to depressive symptoms in primary school children remain insufficiently understood. This study aimed to investigate whether loneliness, perceived social support, and parent-child relationship mediate the association between family SES and depressive symptoms, and whether these pathways are moderated by sex and household registration.

Methods

A three-wave longitudinal survey was conducted among primary school students in Taizhou City, Zhejiang Province, China (2021–2023). Data were collected via online questionnaires completed by students and their parents. Family SES indicators (parents’ education, family economic status, and average living space per person) were assessed at baseline (T1), interpersonal experiences (loneliness, perceived social support, and parent-child relationship) at wave 2 (T2), and depressive symptoms at wave 3 (T3). Structural equation modeling was used to examine direct, mediated, and moderated effects, adjusting for relevant covariates.

Results

A total of 2,164 students (mean age = 9.3 ± 1.4 years, 46.5% female) who completed all three waves were included in the analysis. Higher parents’ education was directly associated with lower depressive symptoms and indirectly associated via reduced loneliness (indirect effect − 0.02, 95% CI -0.03 to -0.001; 13.2% of the total effect) and greater perceived social support (indirect effect − 0.02, 95% CI -0.03 to -0.01; 16.3%). Better family economic status was linked to fewer depressive symptoms both directly and indirectly through greater perceived social support (indirect effect − 0.01, 95% CI -0.02 to -0.001; 11.1%). Larger average living space per person was associated with lower depressive symptoms only indirectly through increased perceived social support (indirect effect − 0.01, 95% CI -0.02 to -0.002; 43.5%). Household registration moderated the indirect pathway via perceived social support: the indirect effect of family economic status was significant among urban children but not among rural children. No significant moderation by sex was observed.

Conclusions

Family SES was associated with depressive symptoms in primary school children partly through loneliness and perceived social support, with the latter pathway contingent on household registration. These findings underscore the importance of interpersonal processes and contextual conditions in shaping socioeconomic inequalities in childhood mental health and highlight the need for context-sensitive prevention strategies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13034-026-01116-2.

Background

Depressive symptoms have been recognized as a major mental health concern in childhood, with prevalence estimates rising globally over recent decades [1, 2]. A recent systematic review indicates that the pooled prevalence of depressive symptoms among Chinese primary school students ranges from approximately 14% to 20% [3], suggesting that emotional problems often emerge well before adolescence. Children in primary school are typically in Piaget’s concrete operational stage (around 7–11 years), a critical developmental period characterized by rapid advances in cognitive abilities, self-concept, and emotion regulation [4]. If not recognized and intervened upon early, depressive symptoms tend to persist, undermining academic performance, health behaviors and social functioning, and increasing the risk of recurrent depression across adolescence and adulthood [5].

From the perspective of ecological systems theory, children’s psychological development is shaped by dynamic interactions across multiple nested systems, with the family serving as the most proximal and influential microsystem [6, 7]. Within this framework, family socioeconomic status (SES) constitutes a key structural determinant of children’s emotional well-being. Family SES influences not only access to material and educational resources but also the quality of family relationships and the broader social environments in which children engage on a daily basis [8]. In the context of rapid socioeconomic changes and pronounced urban-rural disparities in China, clarifying how family SES relates to children’s depressive symptoms, through which mechanisms and under what conditions, is of particular significance.

A substantial body of research has documented that low family SES is associated with elevated internalizing problems, including depression, among children and adolescents [9, 10]. Two complementary theoretical perspectives offer insight into why family SES may be linked to childhood depression. The family stress model posits that family economic hardship generates parental living burden and psychological distress, which in turn disrupts supportive parenting practices and increases the risk of emotional and behavioral problems in children [11, 12]. Conversely, the family investment model emphasizes that socioeconomically advantaged families are better able to invest in cognitively stimulating environments, emotional support and opportunities for extracurricular engagement, thereby fostering resilience and reducing vulnerability to depression [13]. Together, the present evidence and perspectives indicate that family SES may influence children’s depressive symptoms through both direct and indirect pathways, operating via children’s experiences in their interpersonal environments [14].

Children’s interpersonal experiences, particularly loneliness and perceived social support, are likely to be critical pathways through which family SES may shape depressive symptoms. Loneliness, defined as a distressing feeling arising from a perceived deficit in the quality or quantity of one’s social relationships, is a robust predictor of depression across the lifespan [15, 16]. Emerging evidence suggests that lower family SES is associated with higher levels of loneliness in youth, possibly due to constrained access to social activities and supportive networks [17]. Perceived social support, reflecting individuals’ beliefs regarding the availability and adequacy of supportive resources [18], is hypothesized to be another key interpersonal mechanism. SES stratifies access to and perception of social resources, with higher-SES children reporting more supportive friendships, parental involvement, and social integration, whereas lower-SES peers experience diminished support and greater emotional vulnerability [19, 20]. A growing body of evidence indicates that perceived social support functions as a well-established protective factor against depression in childhood [21, 22].

Beyond the broader interpersonal experiences, the parent-child relationship represents a foundational relational context in children’s emotional development. As children’s earliest and most enduring interpersonal bond, the quality of parent-child relationships plays a central role in shaping emotional security and psychological well-being [23]. Prior studies demonstrate that family SES may be associated with the relational climate within the home, with higher-SES families tending to foster more supportive and responsive parent-child interactions [24]. Longitudinal evidence further indicates that poorer parent-child relationship quality is associated with an elevated risk of subsequent depressive symptoms in children and adolescents [25, 26]. Mediation research, largely based on cross-sectional evidence, suggests that associations between family SES and children’s mental health may be partly explained by family relational processes, such as parenting quality and parental engagement, rather than solely by direct pathways [27]. However, given that school-aged children increasingly engage with peers and other adults beyond the family, focusing exclusively on the parent-child relationship may overlook additional interpersonal pathways linking SES to depression. Accordingly, integrating multiple interpersonal experiences across family and broader social contexts is essential for a more comprehensive understanding of SES-related emotional risk.

Importantly, these mediating processes may not operate uniformly across all children, highlighting the need to consider moderating factors that shape for whom and under what conditions these pathways are most pronounced. Evidence suggests that the associations between family SES and children’s depressive symptoms vary across sociodemographic groups, underscoring the value of moderator models in clarifying differential susceptibility [28]. Sex has been widely discussed as a potential source of heterogeneity in children’s depressive symptoms, as girls tend to report higher levels of depressive symptoms and are more sensitive to interpersonal stressors such as conflict and rejection, whereas boys may be more influenced by achievement-related stress [29]. Contextual factors such as household registration may further condition the role of interpersonal resources in children’s mental health. Rather than reflecting a simple rural-urban disadvantage gradient, household registration captures distinct social environments characterized by differences in institutional access, community cohesion, and the availability of informal support networks [30]. These contextual differences may shape how family socioeconomic resources are translated into children’s perceived social support and, in turn, their emotional well-being [31]. Accordingly, examining household registration as a moderator is essential for understanding whether and under what contextual conditions interpersonal resources function as salient pathways linking family SES to depressive symptoms.

Taken together, existing research points to a plausible mediated pathway whereby family SES shapes children’s interpersonal environments, which further influence their risk of depression. However, several important gaps remain. First, most prior studies have focused on adolescents, with relatively limited longitudinal evidence among primary school children, a developmental period during which emotional regulation and social relationships are rapidly evolving. Second, existing research has often examined interpersonal mechanisms in isolation, rather than simultaneously considering multiple interpersonal experiences that span family and broader social contexts [32, 33]. Third, little is known about whether and how these pathways vary across sociodemographic subgroups, particularly with respect to sex and household registration, which may shape the salience and effectiveness of different interpersonal experiences.

To address these gaps, the present longitudinal study of Chinese primary school children aims to: (1) examine the direct association between family SES and children’s depressive symptoms; (2) test whether interpersonal experiences, including loneliness, perceived social support, and parent-child relationship, mediate this association; and (3) explore whether sex and household registration moderate both the direct and indirect pathways. We hypothesize that the association between family SES and children’s depressive symptoms is at least partly mediated by these interpersonal experiences, and that the strength of these pathways varies across social and demographic contexts. By elucidating the underlying mechanisms and conditional patterns of these associations, the findings may inform more precise and targeted prevention strategies.

Methods

Participants

A longitudinal cohort study was conducted among primary school students in Taizhou City, Zhejiang Province, China. Using the multistage cluster sampling, six districts or counties were selected, comprising two urban districts, two county-level cities, and two counties. In each area, three primary schools were selected by simple random sampling, and two classes per grade in each school were randomly cluster chosen. The data collection occurred at three waves: October-November 2021 (T1), April-May 2022 (T2), and February-March 2023 (T3). The details of the participant selection are shown in Fig. S1. Eligibility criteria were met as all follows: (1) enrollment in the sampled classes with the ability to complete online questionnaires under teacher guidance and (2) provision of online informed consent from both students and their parents or legal guardians. Students completed online questionnaires under teacher supervision, while parents or legal guardians completed parallel questionnaires via the Wenjuanxing platform (https://www.wjx.cn), which effectively avoided missing data.

This study was conducted in accordance with all pertinent ethical guidelines and regulations. Informed consent was obtained from all participants and their parents or legal guardians prior to their involvement. The study protocol received formal approval from the Ethics Committee of Taizhou Central Hospital (Approval No.: 2022 L-01‐17).

Measures

Family socioeconomic status

To operationalize family SES, we used three parent-reported indicators, parents’ education, family economic status, and average living space per person. These measures reflect key SES dimensions, including parents’ cultural capital, material resources and living conditions [34]. Because different SES components may have distinct implications for children’s interpersonal environments and mental health, we included these indicators as separate observed variables in the analysis rather than combining them into a single index [35].

1. Parents’ education was measured as a composite indicator reflecting the highest educational level attained by both the father and the mother, based on the approach of Schwartz and Mare [36]. Each parent’s education was assessed using a four-point Likert-type scale ranging from primary school or lower to college or above. The composite score was derived by summing the individual scores of both parents, resulting in a possible range of 2 to 8. Higher scores indicate a higher overall educational level of the parents.

2. Family economic status was assessed using a parent-reported subjective measure similar to those employed in prior research on subjective socioeconomic status [37]. Parents were asked, “How would you describe your family’s economic situation relative to others in your local area?” Responses were initially coded on a three-point Likert-type scale (1 = poor, 2 = fair, 3 = good). The original categories are shown in Table 1. Due to small cell sizes in the extreme categories, the variable was dichotomized (0 = relatively poor; 1 = fair or good) for structural equation modeling analysis to enhance model stability.

3. Average living space per person was calculated by dividing the total household floor area by the number of co-residents and expressed in square meters per person.

Table 1.

Characteristics of participants in the analysis sample

Variables	Overall (n = 2164)	Variables	Overall (n = 2164)
T1		Family economic status
Age, years	9.3 ± 1.4	Good	191 (8.9%)
Sex		Fair	1853 (85.6%)
Female	1006 (46.5%)	Poor	120 (5.5%)
Male	1158 (53.5%)	Average living space per person, m2	32.2 ± 14.8
School type		T2
Public school	1956 (90.4%)	Relationship with father
Non-public school	208 (9.6%)	Good	1943 (89.8%)
Household registration		Normal	200 (9.2%)
Urban	831 (38.4%)	Poor	21 (1.0%)
Rural	1333 (61.6%)	Relationship with mother
Parents’ marital status		Good	2038 (94.2%)
Married	2006 (92.7%)	Normal	114 (5.3%)
Others	158 (7.3%)	Poor	12 (0.6%)
Frequency of physical activity per week	3.4 ± 1.8	Loneliness
Father’s education		Relational connectedness	1.3 ± 0.5
Primary school or lower	68 (3.1%)	Social connectedness	1.2 ± 0.4
Middle school	586 (27.1%)	Self-perceived isolation	1.2 ± 0.4
High school	669 (30.9%)	Social support
College or higher	841 (38.9%)	Family support	6.1 ± 1.0
Mother’s education		Friends support	5.8 ± 1.1
Primary school or lower	83 (3.8%)	Other significant support	5.9 ± 1.0
Middle school	513 (23.7%)	T3
High school	587 (27.1%)	CDI score	5.0 ± 5.6
College or higher	981 (45.3%)	Depressive symptoms	80 (3.7%)

Data are mean (SD) or n (%) unless otherwise indicated. Frequency of physical activity per week, excluding participation in physical education classes

Depressive symptoms

The 27-item Children’s Depression Inventory (CDI) [38] was used to measure depressive symptoms during the last 2 weeks. The CDI has demonstrated strong applicability for individuals aged 7 to 17 years, and its Chinese version has been validated in Chinese children and adolescents [39]. Notably, with low literacy demand, it proves particularly suitable for evaluating depressive symptoms within primary students [40, 41]. Each of the 27 items is scored from 0 to 2 (e.g., “I get sad from time to time; I am sad many times; I am sad all the time”), yielding total scores ranging from 0 to 54. Higher scores signal a more intense tendency toward depressive symptoms. Those who had scores of 19 or above were considered to have depressive symptoms [42]. The scale demonstrated a Cronbach’s alpha coefficient of 0.88 in this study.

Mediating factors

Three variables, namely loneliness, perceived social support and parent-child relationship, were used to reflect the multiple interpersonal experiences of the participants, all of which were self-reported by students.

1. Feelings of loneliness were measured using the (University of California at los Angels, UCLA) Three-Item Loneliness Scale [43]. This instrument assesses three dimensions: relational connectedness, social connectedness and self-perceived isolation. Each item is rated from 1 (hardly ever) to 3 (often), and the total score ranges from 3 to 9, with higher values reflecting greater perceived loneliness. The Chinese version of this scale has demonstrated satisfactory reliability and validity across diverse samples of Chinese adolescents and children [39]. In the current sample, the scale showed acceptable internal consistency, with a Cronbach’s α of 0.78.

2. Perceived social support was evaluated by the Multidimensional Scale of Perceived Social Support (MSPSS) [44], a 12-item instrument designed to measure support from three distinct sources: family, friends and significant others. Each item is rated on a seven-point Likert-type scale, ranging from 1 (very strongly disagree) to 7 (very strongly agree). Scores for each subscale (Family: items 3, 4, 8, 11; Friends: items 6, 7, 9, 12; Significant Other: items 1, 2, 5, 10), as well as the total scale score, were calculated by summing the corresponding items and dividing by the number of items, with higher scores indicating greater perceived social support. Excellent internal consistency was observed for the MSPSS in this study (Cronbach’s α = 0.95).

3. Parent-child relationship was measured with two items (relationship with father and mother) on a 3-point scale (poor = 1, normal = 2, good = 3). The sum score, ranging from 2 to 6, represented overall relationship quality, with higher scores indicating a more positive relationship.

Covariates and moderating factors

All covariates were based on data at T1. Sociodemographic characteristics included age, sex (female, male), school type (public school, non-public school), household registration (urban, rural), and parents’ marital status (married, others). In addition, we incorporated the weekly frequency of physical activity (excluding physical education classes) as a covariate in the model. Except for sex and physical activity, all other variables were reported by parents. Sex and household registration were examined as potential moderating factors. The selection of these moderators was informed by prior literature documenting their associations with family SES, children’s depressive symptoms, and interpersonal experiences [32, 33].

Statistical analysis

For each participant, same data were collected at all three waves (T1, T2, and T3) while were analyzed using a longitudinal framework. In the correlation analysis and multiple mediation model, family SES indicators assessed at T1, including parents’ education, perceived family economic status, and average living space per person, were specified as exposures. Interpersonal factors assessed at T2, including loneliness, perceived social support, and parent-child relationship, were specified as intermediate variables. Depressive symptoms measured at T3 were specified as the outcome. Covariates potentially associated with the mediators and outcome were included in all models. A detailed description of study variables is provided in Table S1 [see Supplementary Material].

Descriptive statistics were used to characterize the sample and study variables. Continuous variables were presented as means and standard deviations, whereas categorical variables were summarized using frequencies and percentages. As an initial analytic step, regression-based analyses were performed to examine associations among family SES indicators, interpersonal factors, and depressive symptoms prior to fitting the full structural equation model. These analyses adjusted for age, sex, school type, household registration, parents’ marital status, and weekly frequency of physical activity. Multicollinearity among three family SES indicators was assessed using variance inflation factors (VIFs). All VIF values were well below the commonly accepted threshold of 5 (range: 1.10–1.33), indicating no evidence of problematic multicollinearity. All statistical tests were two-tailed, and statistical significance was defined as p < 0.05 without adjustment for multiple comparisons.

Primary analyses were conducted using structural equation modeling (SEM) to simultaneously estimate direct and indirect associations between family SES indicators and depressive symptoms through multiple interpersonal mediators. Loneliness, perceived social support, and parent-child relationship were entered concurrently as mediators to account for their shared variance and to reduce potential bias associated with omitted pathways. Standardized regression coefficients were reported to facilitate comparison across pathways. The coefficients were interpreted with regard to direction, magnitude, and statistical significance. The proportion of the total effect explained by each indirect pathway was calculated to quantify the extent of mediation.

Family SES indicators were modeled as separate observed predictors, with parents’ education specified as antecedent to perceived family economic status, which in turn preceded average living space per person. This ordering reflects established theoretical and empirical evidence that parental education represents relatively stable human and cultural capital shaping families’ economic resources, which subsequently influence material living conditions [45, 46]. Correlations rather than directional paths were specified among loneliness, perceived social support, and parent-child relationship. As these interpersonal factors were assessed at the same wave, modeling them as correlated mediators provides a more conservative and theoretically appropriate specification when clear temporal ordering cannot be assumed [47].

In sensitivity analyses, we separately conducted three mediation models with the dimensions or items of each latent mediator added in a single structural equation model. For example, for the mediating role of perceived social support, the three dimensions of family, friends and significant others were tested as mediators in one model. The models were similar to those in the primary analyses except for the mediating variables to evaluate how individual factors of each mediator affected results.

We examined the potential moderating effects of sex and household registration on the predictor-to-mediator path, on the mediator-to-outcome path and on the predictor-to-outcome direct path. This was accomplished by introducing multiplicative interaction terms between the predictors (family SES) and the proposed moderators, as well as between the mediators (interpersonal experiences) and the proposed moderators, into the final structural equation model. For models with a significant interaction term, we computed the index of moderated mediation to quantify how the indirect effect was conditioned on the moderator [48]. Statistical significance was determined by a p value < 0.05 coupled with a bootstrapped 95% confidence interval (CI) that excluded zero [49]. Significant interactions were further probed by computing simple slopes at representative values of the moderator.

All structural equation models were conducted in R (version 4.3.1) using the lavaan package (version 0.6–16). Model fit was evaluated using multiple fit indices, including the root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), comparative fit index (CFI), and Tucker-Lewis index (TLI). RMSEA and SRMR values ≤ 0.06 were considered indicative of good fit, while values up to 0.08 suggested acceptable fit. CFI and TLI values ≥ 0.95 were interpreted as reflecting good fit, and values ≥ 0.90 as acceptable fit [50].

Results

Participant’s characteristics

A total of 2,164 participants who completed all three survey waves from T1 to T3 were included in the current analysis. Table 1 presents the characteristics of participants in the analytic sample. The mean age of participants at baseline (T1) was 9.3 years (SD 1.4). Of the participants, 1006 (46.5%) were female, and 831 (38.4%) had urban household registration. More than 30% of fathers and mothers had attained a college education or higher. Most families reported fair (85.6%) or good (8.9%) economic status. The average living space per person was 32.2 m² (SD 14.8). As shown in Supplementary Table S2, 79 students (3.7%) were classified as having depressive symptoms at baseline based on the established CDI cutoff (≥ 19), and the prevalence of depressive symptoms varied across family SES indicators. The comparison of baseline characteristics between the participants in the study and those who dropped out are presented in Table S3 [see Supplementary Material].

The mediating role of loneliness, perceived social support and parent-child relationship

Using maximum likelihood estimation with 5000 bootstrapped iterations, the final full structural model exhibited good model fit (RMSEA 0.035, sRMR 0.021, CFI 0.975, TLI 0.959). In the preliminary analysis (Table 2) and the final model (Fig. 1; Table 3),

Table 2.

Standardized effects for associations between exposures, mediators, and outcomes

Associations	Standardized coefficients
Parents’ education > loneliness	-0.06 (-0.11, -0.004) *
Family economic status > loneliness	-0.05 (-0.10, -0.003) *
Average living space per person > loneliness	-0.02 (-0.07, 0.03)
Parents’ education > social support	0.12 (0.07, 0.17) ***
Family economic status > social support	0.06 (0.01, 0.10) **
Average living space per person > social support	0.06 (0.01, 0.10) *
Parents’ education > parent-child relationship	0.04 (-0.02, 0.10)
Family economic status > parent-child relationship	0.02 (-0.04, 0.07)
Average living space per person > parent-child relationship	0.02 (-0.03, 0.08)
Parents’ education > depressive symptoms	-0.11 (-0.15, -0.09) ***
Family economic status > depressive symptoms	-0.09 (-0.13, -0.04) ***
Average living space per person > depressive symptoms	-0.02 (-0.07, 0.02)
Loneliness > depressive symptoms	0.38 (0.35, 0.42) ***
Social support > depressive symptoms	-0.35 (-0.39, -0.31) ***
Parent-child relationship > depressive symptoms	-0.21 (-0.26, -0.16) ***

*p value < 0.05, **p value < 0.01, ***p value < 0.001. > indicates the direction of the regression path, that is, parents’ education predicts loneliness

Fig. 1.

Standardized coefficients for the direct effects for the final full structural model. Single-headed arrows indicate regression paths, double‐headed arrows indicate covariances, ovals represent latent variables, and rectangles represent measured variables. Coefficients are shown for statistically significant paths, whereas paths with dashed lines were not significant

Table 3.

Standardized regression coefficients for the direct and indirect effects for the final full structural model

	Family economic status	Average living space per person	Loneliness	Social support	Parent-child relationship	Depressive symptoms
Mediators
Loneliness (direct effect)	—	—	—	—	—	0.28 (0.22, 0.34) ***
Social support (direct effect)	—	—	—	—	—	-0.17 (-0.23, -0.12) ***
Parent-child relationship (direct effect)	—	—	—	—	—	-0.11 (-0.18, -0.03) **
Parents’ education
Direct effect	0.21 (0.17, 0.25) ***	0.22 (0.18, 0.26) ***	-0.06 (-0.11, -0.01) *	0.12 (0.07, 0.17) ***	0.04 (-0.03, 0.11)	-0.06 (-0.11, -0.02) **
Total indirect effects	—	—	—	—	—	-0.07 (-0.09, -0.04) ***
Via loneliness	—	—	—	—	—	-0.02 (-0.03, -0.001) *
Via social support	—	—	—	—	—	-0.02 (-0.03, -0.01) ***
Via parent-child relationship	—	—	—	—	—	-0.004 (-0.01, 0.004)
Via family economic status	—	0.04 (0.03, 0.05) ***	-0.01 (-0.02, 0.00)	0.01 (0.00, 0.02) *	0.00 (-0.01, 0.02)	-0.02 (-0.03, -0.01) **
Via average living space per person	—	—	0.00 (-0.02, 0.01)	0.01 (0.00, 0.02) *	0.00 (-0.01, 0.02)	-0.01 (-0.02, 0.01)
Total effect	—	—	—	—	—	-0.13 (-0.18, -0.08) ***
Family economic status
Direct effect	—	0.19 (0.16, 0.22) ***	-0.05 (-0.11, 0.01)	0.06 (0.01, 0.11) *	0.02 (-0.05, 0.08)	-0.06 (-0.11, -0.01) *
Total indirect effects	—	—	—	—	—	-0.03 (-0.06, -0.01) *
Via loneliness	—	—	—	—	—	-0.02 (-0.03, 0.002)
Via social support	—	—	—	—	—	-0.01 (-0.02, -0.001) *
Via parent-child relationship	—	—	—	—	—	-0.00 (-0.01, 0.01)
Via average living space per person	—	—	0.00 (-0.01, 0.01)	0.01 (0.00, 0.02) *	0.00 (-0.01, 0.01)	-0.002 (-0.01, 0.01)
Total effect	—	—	—	—	—	-0.09 (-0.14, -0.04) *
Average living space per person
Direct effect	—	—	-0.02 (-0.07, 0.04)	0.06 (0.01, 0.11) *	0.02 (-0.03, 0.08)	-0.01 (-0.05, 0.04)
Total indirect effects	—	—	—	—	—	-0.02 (-0.04, 0.00)
Via loneliness	—	—	—	—	—	-0.01 (-0.02, 0.01)
Via social support	—	—	—	—	—	-0.01 (-0.02, -0.00) *
Via parent-child relationship	—	—	—	—	—	-0.00 (-0.01, 0.00)
Total effect	—	—	—	—	—	-0.02 (-0.07, 0.02)
Covariances
Loneliness	—	—	—	-0.45 (-0.50, -0.39) **	-0.18 (-0.26, -0.09) **	—
Social support	—	—	-0.45 (-0.50, -0.39) **	—	0.31 (0.24, 0.38) **	—

*p value < 0.05, **p value < 0.01, ***p value < 0.001

higher parents’ education was directly associated with lower levels of loneliness, greater perceived social support, and less severe depressive symptoms. In the final model, the effect of parents’ education on depressive symptoms was mediated by loneliness (indirect effect − 0.02, 95% CI -0.03 to -0.001) and perceived social support (indirect effect − 0.02, 95% CI -0.03 to -0.01), with longitudinal mediation effect sizes of 13.2% and 16.3%, respectively. Higher parents’ education was also positively associated with family economic status (direct effect 0.21, 95% CI 0.17 to 0.25) and average living space per person (direct effect 0.22, 95% CI 0.18 to 0.26), and was indirectly associated with the mediators and depressive symptoms through these socioeconomic indicators.

In the preliminary analysis (Table 2), better family economic status was directly associated with lower levels of loneliness, higher levels of social support, and less severe depressive symptoms. In the final model (Fig. 1; Table 3), perceived social support mediated the association between family economic status and depressive symptoms (indirect effect − 0.01, 95% CI -0.02 to -0.001), accounting for 11.1% of the total effect. In contrast, the direct association between family economic status and loneliness was no longer statistically significant.

Moreover, better family economic status was positively associated with average living space per person (direct effect 0.19, 95% CI 0.16 to 0.22) and showed indirect associations with perceived social support through average living space per person. The pathway from average living space per person to social support and subsequently to depressive symptoms was observed (indirect effect − 0.01, 95% CI -0.02 to -0.002), accounting for 43.5% of the total effect. No significant direct effect of average living space per person on other mediators or depressive symptoms was found.

Greater loneliness was positively associated with depressive symptoms (direct effect 0.28, 95% CI 0.22 to 0.34). In contrast, greater perceived social support (direct effect − 0.17, 95% CI -0.23 to -0.12) and parent-child relationship (direct effect − 0.11, 95% CI -0.18 to -0.03) were negatively associated with depressive symptoms. In addition, the three mediators were significantly intercorrelated.

In our sensitivity analyses, the results from the single mediator models were largely consistent with the final model estimates (Table S4, see Supplementary Material). Two of three items of the loneliness measure (relational connectedness and self-perceived isolation) mediated the effect of parents’ education on depressive symptoms (Fig. S2, see Supplementary Material). Both parents’ education and average living space per person had indirect effects via family and other significant support on depressive symptoms, whereas family economic status exhibited an indirect association only through family support (Fig. S3, see Supplementary Material). No mediated effect of parent-child relationship was observed between family SESs and depressive symptoms (Fig. S4, see Supplementary Material).

Moderated mediation analysis

We first examined whether sex and household registration moderated the structural pathways in the model. Sex did not statistically moderate any of the predictor-to-mediator, mediator-to-outcome, or direct predictor-to-outcome paths (all p values > 0.05), and the indices of moderated mediation by sex were not significant (95% CIs included zero).

In contrast, a significant moderating effect of household registration (urban vs. rural) was observed only on the pathway from family economic status to perceived social support (interaction coefficient = 0.29, 95% CI 0.03 to 0.56; p = 0.031). As shown in Fig. 2 and Table S5 [see Supplementary Material], the conditional effect of family economic status on perceived social support was significant among children with urban household registration (coefficient = 0.33, 95% CI 0.07 to 0.59, p = 0.012), whereas the corresponding effect was not statistically significant among those with rural household registration (coefficient = 0.04, 95% CI -0.01 to 0.09; p = 0.154). The index of moderated mediation was significant (coefficient = -0.05, 95% CI -0.10 to -0.01, p = 0.042), indicating that household registration moderated the indirect association between family economic status and depressive symptoms through perceived social support. The conditional indirect effect for urban children was significant (indirect effect =-0.06; 95% CI -0.11 to -0.01, p = 0.020), but not among their rural counterparts (indirect effect =-0.01; 95% CI -0.02 to 0.01, p = 0.168).

Fig. 2.

Household registration moderating the mediation of social support on depressive symptoms. *p value < 0.05, **p value < 0.01, ***p value < 0.001. Coefficients are standardized. a₁, the path coefficient from family economic status to social support among children with rural household registration. a₂, the path coefficient from family economic status to social support among children with urban household registration. b, the path coefficient from social support to depressive symptoms (this is shared across household registration groups, as there is no moderating effect in this path). IE₁, indirect effect for rural household registration. IE₂, indirect effect for urban household registration. c, total effects. c’, direct effect of family economic status on depressive symptoms after adjustment for the mediating effects (this is shared across household registration, as there is no moderating effect in this path)

Discussion

This longitudinal study among primary school children in China examined how family SES is associated with depressive symptoms through multiple interpersonal experiences, and whether these pathways vary across sociodemographic contexts. The results indicated that both higher parents’ education and better family economic status were associated with fewer depressive symptoms over time. Loneliness and perceived social support partially mediated these associations, whereas the parent-child relationship did not serve as a significant mediator. Household registration moderated the indirect pathway linking family economic status to depressive symptoms via perceived social support, with a significant indirect effect observed among urban but not rural children. Together, these findings suggest a complex, context-dependent process through which socioeconomic advantage may shape children’s emotional development, partly by reducing loneliness and enhancing perceived social support.

Direct associations between family SES and depressive symptoms

Consistent with prior research [9, 10], we found that higher parents’ education directly predicted lower depressive symptoms over time. Parental educational attainment may reflect not only socioeconomic advantages but also greater awareness of child development, more effective communication strategies, and healthier coping styles, which together help create emotionally supportive environments [45]. Family economic status also showed a direct negative association with depressive symptoms in the final model, suggesting that parents’ perception of economic adequacy may influence children’s emotional health beyond interpersonal processes. Better economic conditions may reduce daily stressors and enhance household stability, thereby lowering children’s vulnerability to negative affect [51]. In our model, the direct association between parents’ education and depressive symptoms was comparable in magnitude to that of family economic status, underscoring the independent relevance of multiple SES dimensions in shaping children’s emotional well-being. By contrast, average living space per person was not directly associated with depressive symptoms, indicating that objective housing conditions may be more distal and exert their influence primarily through more proximal psychosocial mechanisms [52]. Overall, these findings suggest that both parents’ education and perceived economic status are salient SES indicators linked to children’s depressive symptoms, whereas housing space may play a more indirect role.

The mediating role of loneliness, perceived social support and parent-child relationship

Loneliness significantly mediated the longitudinal association between parents’ education and depressive symptoms. Children from more highly educated families may benefit from stronger socioemotional competencies and more supportive communication patterns, which may, in turn, reduce the likelihood of social isolation [53, 54]. Consistent with interpersonal theories of depression [55], our findings showed that reduced loneliness, particularly diminished self-perceived isolation and relational disconnection, subsequently mitigated depressive symptoms. This finding underscores loneliness as a core interpersonal vulnerability for depression during primary school years, a developmental stage characterized by increasing peer orientation and sensitivity to social belonging [56, 57].

Perceived social support emerged as another robust mediator linking multiple family SES indicators to depressive symptoms. In our study, higher parents’ education and better family economic status predicted stronger perceived social support, which in turn was associated with lower depressive symptoms. This pattern aligns with family investment theory and prior evidence showing that socioeconomically advantaged families are better positioned to facilitate access to supportive social networks and extracurricular activities [58, 59]. Notably, average living space per person indirectly predicted depressive symptoms through perceived social support, indicating that material living conditions may influence children’s emotional well-being by enabling social participation and opportunities for relationship building rather than through direct psychological effects [46, 60].

Despite being a strong prospective predictor of depressive symptoms [61], the parent-child relationship did not mediate the association between family SES and depressive symptoms. Several explanations may account for this finding. First, the brief measure used in this study may have captured general relationship quality rather than specific dimensions such as warmth, responsiveness, or conflict, which have been shown to be more sensitive to socioeconomic variation [62, 63]. Second, during the primary school years, parent-child relationship quality may be relatively stable and shaped more by long-term caregiving patterns than by structural socioeconomic conditions, thereby limiting its role as a pathway through which SES disparities are transmitted [45, 64]. In contrast, children’s peer-related experiences and broader social environments may be more immediately responsive to socioeconomic resources and constraints. Importantly, this pattern remained robust in sensitivity analyses in which relationships with fathers and mothers were included simultaneously as mediators, without adjustment for loneliness or perceived social support. Together, these findings highlight that, although the parent-child relationship is an important proximal factor for children’s emotional health, it may not represent a primary mechanism linking family SES to depressive symptoms, underscoring the importance of considering multiple interpersonal domains beyond parent-child interactions.

Moderation by sex and household registration

In the present study, we found no evidence that sex significantly moderated any of the direct or indirect pathways linking family SES to depressive symptoms. This finding contrasts with some prior research in adolescents showing significant gender differences in sensitivity to interpersonal stressors and emotional outcomes [65, 66]. One possible explanation is the relatively young age of our sample (mean age = 9.3 years), as gender disparities in depressive symptoms tend to become more pronounced following pubertal onset [67]. The lack of evidence for sex differences suggests that similar socioeconomic pathways to depressive symptoms for boys and girls during middle childhood. These findings highlight the potential role of developmental stage in shaping heterogeneity in depression risk.

On the other hand, this study found that household registration moderated the pathway linking family economic status to perceived social support. Specifically, higher family economic status predicted greater perceived social support for urban children, whereas the corresponding association was weak and non-significant for rural children. Accordingly, the indirect effect of family economic status on depressive symptoms through perceived social support was significant only for urban children. This moderation may partly reflect structural disparities in social and institutional resources associated with household registration in China. Urban families generally have greater access to organized extracurricular activities, community-based resources, and formal support systems, which may allow economic advantages to be more readily translated into children’s perceived social support [68]. In contrast, rural children may rely more heavily on kinship-based or informal support networks that are less directly shaped by family economic status [69], potentially attenuating the link between economic resources and broader social support experiences. These findings suggest that socioeconomic advantages may be more readily converted into perceived social support in resource-rich contexts, highlighting the importance of contextual opportunity structures, such as the availability of institutionalized social and community resources, in shaping the effectiveness of family resources.

Practical implications of the study

The present findings have several practical implications for the prevention and intervention of depressive symptoms among primary school children, particularly those growing up in socioeconomically disadvantaged contexts. Given that loneliness and perceived social support served as key mediating factors, efforts to reduce childhood depressive symptoms should prioritize improving children’s interpersonal experiences alongside addressing material and structural conditions. School-based initiatives that foster peer inclusion, social connectedness, and supportive classroom climates may be particularly effective for children from lower-SES backgrounds, who may be more vulnerable to loneliness and limited access to supportive social networks [70]. Such school-based programs represent a promising and scalable approach during primary school, a developmental period in which peer relationships become increasingly central to children’s emotional well-being [71]. In addition, programs that promote participation in extracurricular activities and community engagement may strengthen children’s perceived access to social support, thereby buffering the emotional risks associated with lower family socioeconomic status [72]. The moderating role of household registration highlights the need for context-sensitive intervention strategies. Although the direct association between family economic status and depressive symptoms did not differ by household registration, the conversion of family economic resources into perceived social support varied across contexts. In urban settings, where family economic resources appear more readily convertible into perceived social support, programs that reduce economic barriers to participation in organized social and extracurricular activities may help mitigate socioeconomic disparities in emotional well-being [73]. In contrast, in rural contexts, where family economic resources appear less consistently linked to children’s perceived social support, interventions may need to place greater emphasis on strengthening inclusive school environments and community-based informal support systems, rather than relying solely on family economic resources.

Strengths and limitations

This study has several notable strengths. The multi-wave longitudinal design enabled the examination of temporal associations between family socioeconomic status, interpersonal experiences, and depressive symptoms. The simultaneous consideration of multiple interpersonal mediators allowed for a more comprehensive assessment of the mechanisms linking family SES to children’s emotional health. By explicitly examining whether these pathways varied across sociodemographic contexts, the study further provides a nuanced understanding of how socioeconomic influences operate under different structural conditions. In addition, the use of structural equation modeling with latent factors reduced measurement error, and sensitivity analyses supported the robustness of the main findings.

Nonetheless, several limitations warrant acknowledgment. First, as an observational study, while temporally ordered data were utilized and some evidences on the causal associations between family SES and childhood depressive symptoms were provided in our study, the potential for unmeasured confounding and reverse causality remains. Thus, future research employing more rigorous causal methodologies is warranted to strengthen the causal inference. Second, the study was conducted in a single city, Taizhou, in Zhejiang Province, China, which is relatively economically developed. Therefore, the caution is warranted when extrapolating these findings to other populations, especially in the less developed areas. Third, although psychosocial variables were assessed via child self-report, family socioeconomic status was reported by parents, limiting the risk of shared method variance across key associations. The use of latent factors further helped reduce measurement error. In addition, the relatively broad measure of the parent-child relationship measure may not have captured more nuanced family processes. Future research would benefit from employing more fine-grained measurement tools, and from using experimental or intervention designs, particularly those targeting perceived social support and loneliness, to further validate the mechanisms identified in the present study. Finally, attrition across survey waves represents a potential limitation. Although follow-up rates did not reach ideal levels in this study, such degrees of attrition were similar to those reported in large-scale, school-based longitudinal studies [74]. Supplementary materials show that the current sample and those lost to follow-up did not differ substantially on major baseline characteristics, suggesting that systematic bias may be limited, although results should still be interpreted with caution.

Conclusions

Based on a longitudinal moderated mediation analysis among Chinese primary school children, this study revealed that the association between family SES and depressive symptoms was primarily mediated by loneliness and perceived social support. Furthermore, household registration moderated the indirect pathway via perceived social support, with the effect being significant only for urban-registered children. These findings highlight the importance of active interpersonal experiences in shaping the emotional well-being of school-aged children and underscore the context-dependent nature of how family economic resources are translated into psychosocial benefits. Interventions aimed at reducing loneliness and enhancing social support may be particularly effective in mitigating SES-related disparities in childhood depression, while emphasizing the need for context-sensitive strategies that account for differences in structural and social environments.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1. ^{(242.6KB, docx)}

Abbreviations

SES

Socioeconomic status

CI

Confidence interval

SD

Standard deviation

CDI

Children’s depression inventory

VIF

Variance inflation factor

UCLA

University of California at los Angels

MSPSS

Multidimensional scale of perceived social support

SEM

Structural equation modeling

RMSEA

Root mean square error of approximation

SRMR

Standardized root mean square residual

CFI

Comparative fit index

TLI

Tucker-Lewis index

Author contributions

Haifan Bai, Tingting Wang, and Yuting Yang: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing-Original draft. Haijiang Lin, Xiaoxiao Chen, and Yixuan Liu: Investigation, Data curation, Writing - Review & editing. Jingyi Wang and Chaowei Fu: Conceptualization, Formal analysis, Methodology, Project administration, Writing - Review & editing. All authors reviewed and approved the final version of the manuscript.

Funding

Chaowei Fu received funding from the Brain Science and Brain-like Intelligence Technology-National Science and Technology Major Project (grant number 2022ZD 0209100) and was funded by the General Project of Shanghai Municipal Health Commission (grant number 202240115).Jingyi Wang received funding from the China Medical Board (grant number #22472) ). Haijiang Lin and Xiaoxiao Chen were sponsored by the Special Support Program for High Level Talents in Taizhou (grant number TZ2022-2). The funders had no involvement in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The authors declare that they have no conflicts of interest.

Data availability

Data and materials are available on reasonable request from the corresponding author.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the ethical guidelines set forth in the Declaration of Helsinki. Ethical approval was obtained from the Ethics Committee of Taizhou Central Hospital (2022 L-01-17). Prior to participation, all participants, as well as the parents or legal guardians of minors, provided informed consent. The informed consent process was strictly followed for the collection of all study data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.