After-school homework is a vital component of school education, often considered an effective tool for consolidating classroom knowledge and improving academic performance. In recent years, however, educational research has shifted its focus toward examining the impact of after-school homework on students’ non-cognitive abilities, such as self-efficacy, perseverance, time management, and emotional regulation. Non-cognitive abilities play a crucial role in students’ long-term development, contributing not only to academic success but also to career achievements and social adaptation. A well-balanced approach to after-school homework duration can offer students additional learning opportunities while simultaneously fostering the development of these essential skills (Cooper et al., 2006). Studies have shown that after-school homework requires students to complete tasks within a set timeframe, which helps cultivate time management skills. Zimmerman emphasized that students need to plan their time, set priorities, and monitor progress while completing homework, behaviors that enhance self-regulation abilities. These skills are valuable not only in academics but also in future career development. Additionally, by completing homework, students consolidate classroom knowledge and experience a sense of accomplishment, thereby boosting their self-efficacy (Zimmerman, 2002). Bandura highlighted that self-efficacy is a key factor influencing behavior choices, effort levels, and persistence. A moderate amount of homework offers students opportunities for practice and feedback, helping to strengthen their confidence in their abilities (Bandura, 1997). Moreover, challenging homework tasks foster grit, as perseverance is a critical predictor of long-term success. By overcoming obstacles and remaining focused on complex assignments, students develop persistence (Duckworth and Seligman, 2005). While moderate homework has a positive effect on the development of non-cognitive abilities, excessive homework can negatively affect students’ emotional wellbeing and motivation. Cooper found that excessive homework can lead to increased stress, sleep deprivation, and decreased interest in learning, which in turn undermines the development of non-cognitive abilities. Furthermore, while moderate homework helps develop time management skills and responsibility, excessive homework can result in anxiety and lower self-esteem (Cooper et al., 2006). Pope also noted that high-intensity homework may reduce adolescents’ social and rest time, hindering their social–emotional development (Pope, 2001).

Based on these observations, this study proposes the first research hypothesis.

In recent years, extracurricular academic training (such as tutoring and subject-specific coaching) has experienced rapid growth, particularly in East Asia and some Western countries. While this type of training is primarily designed to improve academic performance, its impact on students’ non-cognitive abilities (such as self-efficacy, stress resilience, and social skills) has increasingly become a focus of academic research. Some studies suggest that moderate levels of extracurricular training can help students develop more efficient learning strategies, thereby enhancing self-efficacy (Bandura, 1997). For example, in math tutoring, setting incremental goals and providing feedback on problem-solving can help students gradually build confidence in their ability to tackle challenging problems. Additionally, personalized guidance from instructors may reinforce goal-oriented behaviors (Zimmerman, 2002). Some extracurricular programs, particularly those that involve repeated practice and mock exams, may help students adapt to high-pressure environments and develop greater stress resilience. However, “over-training” may lead to emotional exhaustion, goal-oriented training may undermine students’ intrinsic motivation for learning, leading to a decline in interest (Deci and Ryan, 2000; Suter, 2016). Larson emphasized that adolescents require unstructured activities (such as free play or club participation) to develop social skills. However, excessive extracurricular training can reduce the time available for such activities, limiting opportunities for peer interaction (Larson, 2001).

Therefore, this study proposes the second research hypothesis.

After-school assignments and off-campus academic tutoring are the two primary forms of extracurricular learning, both extensively studied for their impact on academic performance. However, the influence of these activities on non-cognitive abilities—such as self-efficacy, perseverance, and emotional regulation—is highly complex and significantly moderated by socio-economic status (SES). SES profoundly affects students’ participation in, and outcomes from, extracurricular learning through mechanisms such as resource distribution, cultural capital, and family support. Regarding resource support, empirical studies show that high-SES families can offer quality resources, such as quiet study spaces, smart learning devices, and personalized tutoring, which help students efficiently complete tasks and reduce learning stress (Zhao, 2024). In contrast, low-SES families, due to limited resources, may face environmental distractions and insufficient support, potentially leading to suboptimal learning outcomes during homework and tutoring (Sirin, 2005). In terms of parental involvement, high-SES parents tend to use a “Guided Participation” approach, helping students understand learning goals through questioning and discussion (Lareau, 2018). This method aids in developing students’ critical thinking and intrinsic motivation, indirectly strengthening their non-cognitive abilities (Zhao, 2024). By comparison, low-SES parents, constrained by time or knowledge, are more likely to engage in supervisory involvement, which may unintentionally trigger parent–child conflicts when they cannot provide effective assistance, undermining the positive impact of extracurricular learning (Hoover-Dempsey et al., 2001). Additionally, high-SES families often incorporate extracurricular learning into educational planning through “cultural matching,” viewing tutoring as part of the “elite development path” and thereby reducing students’ anxiety regarding their goals (Bourdieu, 1986).

Based on this analysis, this study proposes the third research hypothesis.

In this study, we used the 2014–2015 academic year database from the China Education Panel Survey (CEPS), a large-scale, longitudinal survey project that provides a representative analysis of the influence of family, school, community, and macro-social structures on individual educational outcomes. The survey also explores how educational outcomes shape an individual’s life trajectory. This project was designed and conducted by the China National Survey Research Center (NSRC) at Renmin University of China.

The China Education Panel Survey (CEPS), launched with the 2013–2014 academic year as its baseline, begins with two cohorts: the first-grade students in junior high school (Seventh Grade) and the third-grade students in junior high school (Ninth Grade). Stratified by average education levels and the proportion of mobile populations, the survey randomly selected 28 county-level units (including counties, districts, and cities) across China as survey sites. The survey was school-based, and within these selected units, 112 schools and 438 classes were randomly chosen for participation. All students in the selected classes were included in the baseline survey, totaling approximately 20,000 students. This project uses questionnaires as its main data collection tool. Surveys are administered to all participating students, their parents or guardians, class and subject teachers, and school principals. The questionnaire covers a broad range of topics, including students’ basic information, life experiences, physical and mental health, parent–child interactions, in-school learning, extracurricular activities, relationships with teachers and classmates, social development, and family member details. It also gathers parents’ basic information, lifestyles, family educational environment, parental investment in education, community environment, views on school education, interactions with teachers, and expectations for their children’s education. In addition, comprehensive cognitive ability tests and basic personality assessments are conducted for students, and key examination results (such as mid-term exams, high school entrance exams, and college entrance exams) are collected. Plans are also in place to conduct health and physical examinations, collect biomedical indicators, and employ various methods and technologies to comprehensively gather high-quality data.

Official reports for the CEPS, such as results, assessment frameworks, user guides, background survey questionnaires, databases, and relevant information, are available on the official website: http://www.cnsda.org/index.php?r=projects/view&id=61662993.

Multilevel linear models are well-suited for hierarchical analyses, as they allow for the effective use of information at different levels. These models decompose the factors influencing students’ non-cognitive abilities across various relevant levels, resulting in more accurate estimates and more meaningful interpretations. Developed to overcome the limitations of traditional statistical techniques in handling multilevel structured data, multilevel linear models prevent potential misinterpretation of results. They are ideal for conducting thorough and appropriate analyses of commonly encountered nested data (Osborne and Neupert, 2013).

In analyzing the impact of post-homework time on students’ non-cognitive abilities, a multilevel linear model is employed. This approach is necessary because, in the CEPS, student samples are nested within different schools, and variations in teaching quality and student-teacher ratios exist across schools. To account for the nested structure of the data, a two-level linear model (HLM2) is used to examine the effects of both student-level and school-level variables on students’ non-cognitive abilities. The model construction begins with estimating the null model to determine the percentage of variance in the dependent variable explained by between-group differences and assess its significance. HLM2 is then constructed based on these initial findings.

Based on the theories of overlearning effect and cognitive load theory, and to verify whether the impact of after-school Learning time on non-cognitive abilities in adolescents follows an inverted “U”-shaped curve, the model includes squared terms for homework time and extracurricular tutoring time. Here, ST²_i represents the school homework duration variable, ET²_i represents the extracurricular tutoring time variable. The quadratic model is constructed as follows: (5) Student − level : Y i = β 0 j + β 1 T i + β 2 S T 2 i + β 3 E T 2 i + β 4 X i + β 5 F i + ε i , ε i ~ N 0 δ 2 (6) School − level : β 0 i = γ 00 + γ 01 S i + μ 0 j , μ 0 j ~ N 0 τ 00

In this study, we used IBM SPSS Statistics 26 for data processing, statistical analysis, and model estimation. All regression analyses, including grouped regressions and robustness checks, were conducted in SPSS 26, ensuring consistency and reproducibility in data processing. To examine the factors influencing students’ non-cognitive ability development concerning after-school Learning time, we employed a multilevel linear model. Based on the null models (Equation 1) and (Equation 2), we first constructed an null model without explanatory variables to analyze the sources of variance in non-cognitive ability development. As shown in Table 2, the intraclass correlation coefficients (ICCs) for Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism are 0.103, 0.134, 0.297, 0.167, and 0.113, respectively. The ICCs exceed the threshold of 0.059, supporting the application of a multilevel linear model for analysis.

Based on the full models (Equation 3) and (Equation 4) as well as the quadratic models (Equation 5) and (Equation 6), we constructed hierarchical linear models to analyze the impact of after-school learning time on adolescents’ non-cognitive ability development. The specific results are shown in Table 3. First, the duration of school homework has a significant negative impact on Neuroticism, while significantly promoting the development of other non-cognitive abilities. All results are significant at the 0.01 and 0.001 levels, indicating that increasing school homework duration suppresses Neuroticism and fosters the positive development of non-cognitive traits such as Openness, Conscientiousness, Extraversion, and Agreeableness. Second, the duration of extracurricular academic tutoring positively affects Openness, Conscientiousness, Extraversion, and Agreeableness, all at the 0.001 level, with no effect on Neuroticism. These findings suggest that increasing the duration of extracurricular tutoring enhances the positive development of adolescents’ non-cognitive abilities in these areas.

Table 3 also show that the quadratic term of school homework duration significantly negatively affects adolescents’ Openness, Conscientiousness, Extraversion, and Agreeableness in both cognitive and non-cognitive abilities, while having a significant positive effect on Neuroticism. These findings suggest that the relationship between school homework duration and the development of adolescents’ cognitive and non-cognitive abilities follows a non-linear pattern. Specifically, the effect of school homework duration on Neuroticism is U-shaped, meaning that as the amount of homework increases, the negative impact on Neuroticism reaches a minimum point, after which the effect weakens. Additionally, it can be inferred that homework duration follows an inverted U-shaped effect on the other four non-cognitive dimensions, suggesting that a moderate amount of school homework promotes the development of Openness, Conscientiousness, Extraversion, and Agreeableness. However, beyond a certain threshold, excessive homework duration has a negative impact. Furthermore, the results show that the quadratic term of extracurricular subject tutoring duration significantly negatively affects adolescents’ Openness, Conscientiousness, Extraversion, and Agreeableness, with no effect on Neuroticism. It is speculated that extracurricular tutoring time has an inverted U-shaped effect on adolescents’ non-cognitive abilities: a certain duration of tutoring promotes the development of Openness, Conscientiousness, Extraversion, and Agreeableness, but exceeding a threshold result in a negative impact. Hence, H1 and H2 is verified.

To further validate the inverted U-shaped relationship, a heterogeneous analysis was conducted by dividing the sample into three groups based on students’ family socioeconomic status (SES): low SES, middle SES, and high SES. The estimation results in Table 4 show that the effect of after-school Learning time on students’ non-cognitive abilities across different socioeconomic backgrounds is consistent with the overall findings presented in Table 3. Specifically, for all SES groups, the duration of school homework exhibits a U-shaped effect on adolescents’ Neuroticism and an inverted U-shaped effect on Openness, Conscientiousness, Extraversion, and Agreeableness. The duration of extracurricular subject tutoring has no significant effect on Neuroticism but demonstrates an inverted U-shaped effect on Openness, Conscientiousness, Extraversion, and Agreeableness. These results further confirm the existence of a critical point in the effect of after-school Learning time on adolescents’ non-cognitive abilities, which optimizes their non-cognitive performance. Hence, H3 is verified.

This study employs the instrumental variable approach to address endogeneity concerns, including reverse causality, in the empirical regression analyzing how socioeconomic status (SES) moderates the effect of after-school learning duration on adolescents’ non-cognitive development.

The instrumental variable for socioeconomic status (SES) is derived from the CEPS parent questionnaire, specifically the item “whether the family receives a subsistence allowance.” Family eligibility for a subsistence allowance is correlated with SES, while having minimal direct influence on adolescents’ non-cognitive abilities, supporting its validity as an instrumental variable. Table 5 presents the first-stage regression results of the instrumental variable “whether the family receives a subsistence allowance” (subsistence allowance) on openness, conscientiousness, extraversion, agreeableness, and neuroticism. The estimated coefficient of the instrumental variable is positive and statistically significant. The Kleibergen-Paap rk LM and Kleibergen-Paap rk Wald F statistics confirm that the instrumental variable (subsistence allowance) successfully passes both the under-identification and weak instrument tests, verifying its effectiveness. The second-stage regression incorporating the instrumental variable demonstrates that SES significantly influences multiple dimensions of adolescents’ non-cognitive abilities. These results indicate that the empirical findings remain robust and credible even after addressing endogeneity concerns.

The estimation results in Tables 3, 4 indicate that the relationship between after-school Learning time and adolescents’ non-cognitive abilities—openness, conscientiousness, extraversion, agreeableness—follows an inverted U-shaped curve, except for neuroticism, which exhibits a U-shaped curve. These findings suggest that longer after-school Learning time does not necessarily result in better development of non-cognitive abilities. Instead, there is a turning point at which the positive effects of after-school Learning time on non-cognitive development begin to decline. The optimal duration of after-school Learning time for adolescents can be determined by calculating the inflection point of the quadratic function in each model.

The results in Table 6 show that, for the full sample of students, a moderate duration of after-school homework promotes the development of non-cognitive abilities such as openness, conscientiousness, extraversion, and agreeableness while reducing neuroticism. However, exceeding a critical threshold reverses these benefits, negatively affecting adolescents’ non-cognitive abilities. For school homework, the optimal durations for promoting adolescents’ non-cognitive abilities are 19.25 min for openness, 43.33 min for conscientiousness, 26.09 min for extraversion, 68.75 min for agreeableness, and 39.23 min for neuroticism. Homework within these time thresholds positively contributes to the development of adolescents’ non-cognitive abilities. For extracurricular subject tutoring, the optimal durations are 107.75 min for openness, 65.71 min for conscientiousness, 92.61 min for extraversion, and 53.25 min for agreeableness. Tutoring within these time frames is more beneficial for fostering the development of these dimensions of adolescents’ non-cognitive abilities. Moreover, considering openness, conscientiousness, extraversion, agreeableness, and neuroticism as five interconnected dimensions of non-cognitive abilities, the average values of these indicators can be used to estimate the optimal durations for after-school learning. The ideal time for school homework is 39.33 min, while the optimal duration for tutoring is 79.83 min.

Furthermore, the critical points for optimal after-school Learning duration vary across different student groups. Teenagers from low SES families have longer optimal after-school learning durations, both for school homework and extracurricular tutoring, compared to those from middle and high SES families. Specifically, for teenagers from low SES families, the greatest developmental effect on non-cognitive abilities occurs when the school homework duration is less than 49.82 min, while extracurricular academic tutoring has the greatest impact when limited to 134.96 min. For teenagers from middle SES families, a school homework duration of 32.16 min is most beneficial for the development of non-cognitive abilities, while extracurricular academic tutoring lasting up to 101.61 min is most effective. To optimize the effect of after-school learning on the development of non-cognitive abilities, the ideal school homework duration for teenagers from high SES families is 15.11 min, with an optimal duration for extracurricular tutoring of 79.42 min.