Effect of parental migration on the noncognitive abilities of left-behind school-going children in rural China

The study investigates how different parental migration statuses affect the non-cognitive abilities of left-behind school-going children (LBCs) in rural China. Against a backdrop of large-scale rural-to-urban migration in China and concerns that only a small fraction of migrant workers relocate with their families, the paper frames a key question in human capital development: does parental migration enhance or harm the non-cognitive development of children who remain in rural homes? Non-cognitive skills (often conceptualized by the Big Five framework) are increasingly recognized as critical determinants of educational success and life outcomes. Prior literature has focused on academic performance, physical and mental health, but robust causal evidence on non-cognitive outcomes remains limited, particularly for rural school-aged children and by parental gender. The paper aims to fill these gaps by evaluating the effects of paternal-only, maternal-only, and both-parents migration on children’s non-cognitive abilities and by addressing selection bias using PSM.

Theoretically, human capital models increasingly incorporate non-cognitive abilities—such as conscientiousness, agreeableness, and emotional stability—as important determinants of educational attainment and labor market success (Heckman et al.; Kautz et al.). Families play a central role in shaping these traits through presence, guidance, and investments. Empirically, studies have linked parental migration to worse educational and psychosocial outcomes among left-behind children in China and elsewhere, though findings vary. Research documents mental health issues and lowered academic achievement among LBCs (e.g., Tong et al., Zhang et al., Meng and Yamauchi), and mixed evidence on non-cognitive effects, with some studies showing negative impacts of parental absence, particularly maternal migration, and others reporting nuanced or context-dependent results (Lee and Park; Wu and Zhang; Zhao et al.; Liu et al.). Methodological limitations—cross-sectional designs, selection bias, and limited measurement—constrain causal inference. The present study contributes by focusing on rural school-aged children, distinguishing paternal vs. maternal vs. dual-parent migration, using composite non-cognitive indices mapped to Big Five traits in CEPS, and applying PSM to mitigate selection bias.

Data: The study uses wave 1 (2013/2014) of the China Education Panel Survey (CEPS), a nationally representative survey of Grades 7 and 9 across 31 mainland provincial-level units, covering 28 county-level units, 112 schools, 438 classes, and 19,487 students. The analytic sample restricts to rural students (N = 4,102). Wave II was not used as the study is not a period comparison or policy intervention analysis. Measures: Dependent variable: Non-cognitive abilities measured via CEPS items aligned with Big Five sub-dimensions: Emotional stability (5 items on mood in past 7 days; 5-point frequency scale), Agreeableness (3 items on school belonging and participation; 4-point agreement scale), and Conscientiousness (3 items on persistence and diligence for school and homework; 4-point agreement scale). A composite non-cognitive score sums these components. Independent variable: Parental migration status with four categories reported by students: both parents at home; only father migrates; only mother migrates; both parents migrate. For PSM, treatment is defined separately for each migration status vs. the control of both parents at home (binary D ∈ {0,1}). Controls: Gender, age, only child status, ethnic minority, presence of an irresponsible/bad companion friend, parental education (mother and father), family financial condition (self-reported scale 1–5), and home computer/internet access (0–2). Model specification: A human-capital production function: Y_is = β0 + β1 MIG_is + β2 X_is + ε_is, where Y_is is the child’s non-cognitive ability; MIG is the migration-status indicator; X_is are controls. The a priori expectation is negative effects of parental migration on non-cognitive abilities. Estimation: Propensity Score Matching (PSM) estimates the Average Treatment effect on the Treated (ATT) for each migration status vs. control (children living with both parents). Steps included: estimating propensity scores P(D=1|X); ensuring common support; covariate balance diagnostics (standardized bias, t-tests); and computation of ATT for the composite and sub-indices. Balance tests showed substantial reduction in Pseudo-R2 and LR chi-square post-matching, low standardized biases across covariates, and adequate common support, satisfying parallel trend and overlap assumptions. Descriptive sample characteristics: Of 4,102 rural students, migration breakdown is: both parents at home = 3,026; only father migrates = 399; only mother migrates = 143; both parents migrate = 534. Non-cognitive composite: mean 32.36 (SD 4.43). Emotional stability: mean 10.02 (SD 3.71). Agreeableness: mean 12.00 (SD 2.80). Conscientiousness: mean 10.34 (SD 1.76).

- Overall, parental migration is associated with detrimental non-cognitive outcomes for left-behind rural students, consistent with the study’s a priori expectation. - Paternal migration (Table 7): ATT on composite non-cognitive ability = -0.965 (SE 0.393, t = -2.45, p < 0.05); Emotional stability: +0.118 (ns); Agreeableness: -0.832 (SE 0.247, t = -3.37, p < 0.01); Conscientiousness: -0.251 (SE 0.148, t = -1.69, p < 0.10). Interpretation: significant declines in overall non-cognitive ability, agreeableness, and marginal decline in conscientiousness. - Maternal migration (Table 8): Composite = +0.636 (ns); Emotional stability: +1.224 (SE 0.568, t = 2.16, p < 0.05); Agreeableness: -0.126 (ns); Conscientiousness: -0.462 (SE 0.224, t = -2.06, p < 0.05). Interpretation: while emotional stability appears higher, conscientiousness significantly declines; authors emphasize net detrimental implications for non-cognitive development. - Both parents migrate (Table 9): Composite = -0.210 (ns); Emotional stability: +0.867 (SE 0.327, t = 2.65, p < 0.01); Agreeableness: -0.919 (SE 0.240, t = -3.83, p < 0.01); Conscientiousness: -0.157 (ns). Interpretation: significant drop in agreeableness notwithstanding a rise in emotional stability; authors conclude overall adverse effects on non-cognitive abilities. - Balance diagnostics (Tables 4–6, Figures 1–2): Post-matching, standardized biases reduced to low single digits; Pseudo-R2 and LR chi-square sharply decreased and became insignificant, and common support was adequate across models. - Policy implication: Findings suggest that separation due to parental migration reduces supervision, guidance, and emotional support, increasing risks for poorer non-cognitive outcomes among LBCs.

The results directly address the research question by showing that paternal-only, maternal-only, and both-parents migration are associated with unfavorable changes in key non-cognitive dimensions among rural left-behind students. Declines in agreeableness are consistent and pronounced, with additional evidence of reduced conscientiousness in paternal- and maternal-migration cases. Although emotional stability shows positive associations in some migration scenarios, the authors argue the broader pattern remains detrimental for overall non-cognitive development, aligning with prior literature on the negative psychosocial consequences of parental absence. Mechanistically, reduced parental oversight, disrupted routines, and diminished emotional support likely impair social and behavioral competencies essential for schooling and longer-term human capital accumulation. These findings underscore the need for policies that facilitate family co-migration or provide targeted support systems for LBCs to mitigate developmental risks and reduce rural-urban disparities.

Parental migration—whether by fathers, mothers, or both—undermines the non-cognitive development of rural left-behind school-going children in China, with notable declines in agreeableness and evidence of reduced conscientiousness under certain migration patterns. The study contributes by distinguishing migration by parental gender and dual absence, employing comprehensive non-cognitive indices from CEPS, and using PSM to address selection bias. Policy recommendations include encouraging family unity in migration decisions, improving rural living conditions and educational infrastructure, reducing educational barriers (e.g., expanding schools for LBCs, lowering tuition), strengthening welfare and supervision systems for LBCs, and promoting balanced economic development to reduce the necessity of long-distance parental migration. Future research should deepen causal understanding of mechanisms and refine measurement of non-cognitive subcomponents.

The CEPS instrument, while rich for educational outcomes, has measurement limitations for certain non-cognitive sub-items; further refinement of questionnaires is needed. The analysis relies on cross-sectional wave 1 data; despite PSM mitigating selection bias, causal interpretation remains constrained relative to longitudinal or experimental designs. Additionally, existing surveys lack detailed information on mechanisms explaining why and how parental migration adversely affects non-cognitive development, indicating a need for targeted data collection.