Integrated macro and micro analyses of student burden reduction policies in China: call for a collaborative "family-school-society" model

China's student burden reduction policy is a key response to UNESCO's SDG 4, seeking to promote quality education by reducing inequities, improving classroom teaching quality, and expanding in-school after-school services. The policies aim to move away from narrow evaluation regimes (e.g., “only grades,” “only admissions”) toward comprehensive development of morality, intelligence, physique, aesthetics, and labor, and to reduce homework and off-campus training burdens (“double reduction”) while supporting children’s all-round development and lifelong education. A strong cultural context of high educational desire among Chinese parents contributes to excessive academic pressure and anxiety, driving over-investment in shadow education and competition for scarce perceived opportunities. Building a "family-school-society" sustainable education system is proposed to address student burden through coordinated roles and responsibilities across stakeholders. Historically, since 1949, burden reduction has progressed through three stages: teaching reform (1955–1999), checklist-based burden reduction (2000–2020), and root-based burden reduction (2021–present). The 2021 Double Reduction policy marked a new stage emphasizing high-quality education, governance of off-campus training, and alleviating anxiety. However, research has rarely integrated macro policy evolution with micro-level parent perspectives. This study asks: (1) What are the development trends and implementation patterns of burden reduction policies? (2) What are parents’ attitudes toward these policies? (3) What is the relation between students’ burden and parents’ attitudes? (4) How can a sustainable system be built to alleviate learning burden? The study analyzes 232 policies (1951–2021) and survey responses from 23,567 parents across 29 provinces.

Design: Two-step mixed analysis integrating macro policy text analysis and micro-level parent survey. Macro-level policy analysis: Collected 232 student burden reduction policy documents from 1951–2021 (62 national, 170 local; ~650,000 words including policies, Q&A, attachments). Conducted word frequency and co-occurrence analyses using Microsoft Excel and Micro World cloud software. Thresholding identified 10,440 words at frequency ≥1 and 232 words at ≥59. Constructed relational networks where node size reflects frequency. Tracked issuing agencies over time and central vs. local issuance patterns. Micro-level parent survey: Online survey (December 2021) across 29 provinces using convenience sampling (notably Guangxi, Inner Mongolia, Sichuan, Hebei, Jilin, Yunnan). Distributed and received 23,567 valid questionnaires via WeChat. Respondent characteristics: 6516 male (27.65%), 17,051 female (72.35%); household registration: 15,601 agricultural (66.2%), 7499 non-agricultural (31.82%), 467 other (1.97%, later cleaned). Support for policy: overall 84% support, 16% not support (after recoding). The “other households” category was used to test recognition and later removed in analysis. Variables: - Independent variable: Parents’ views on Double Reduction (originally 5-point scale; mid-category removed; collapsed to binary: support=0, not support=1). - Dependent variables: Parent-evaluated student burden level; daily homework duration (last semester, this semester; 5-point scale: <30 min, 0.5–1 h, 1–1.5 h, 1.5–2 h, >2 h); sleep duration (last semester, this semester; 6-point scale: ≤6 h, 6–7 h, 7–8 h, 8–9 h, 9–10 h, ≥10 h). - Controls: Parent gender, household type (urban/rural Hukou), education level, occupation/work unit. Analytical strategy: - Partial correlation and linear regression: Assessed associations between parental support and dependent variables, controlling for gender, location, education, occupation. Reported models: y1 (burden level) = 1.442 + 0.132x; F=308.822, t=17.573, p<0.01 (significant). y2 (homework last semester) = 1.864 − 0.017x; F=15.194, t=3.898, p<0.01 (significant). y3 (homework this semester) = 1.84 − 0.001x; F=4.696, t=2.167, 0.05>p>0.01 (not significant at 99% CI). y4 (sleep last semester) = 1.701 + 0.032x; F=37.170, t=6.097, p<0.01 (significant). y5 (sleep this semester) = 1.819 − 0.001x; F=0.021, t=−0.146, p>0.05 (not significant). Where x is parental view (support=0, not support=1). - Common method bias: Harman single-factor test; 32 factors with eigenvalue>1; first factor explained 8.238% (<40% threshold), indicating no serious bias. - Binary logistic regression: Outcome: parental support (support=0, non-support=1). Explanatory variables: gender (male=0, female=1), household type (agricultural=0, non-agricultural=1), education level, occupation. Model fit (Hosmer–Lemeshow Sig.=0.163>0.05) indicated good fit. Reported logit: logit(p)=0.111 + 1.477x1 + 0.818x2 + 1.054x3 + 0.998x4 (coefficients reported as log-odds; interpretation presented in text as relative likelihood). - Robustness/endogeneity: Propensity score matching (nearest neighbor, caliper 0.1) in SPSS; 1726 pairs matched; 3199 exact matches; 3084 fuzzy matches. Post-PSM logistic regression maintained good fit (Hosmer–Lemeshow Sig.=0.196>0.05). Post-PSM logit: logit(p)=0.33 + 1.577x1 + 1.199x2 + 1.029x3 + 1.002x4. Coefficient patterns remained qualitatively stable, supporting robustness. Ethics: Approved by Ethical Review Committee, Faculty of Education, Beijing Normal University; informed consent obtained.

- Policy evolution and volume: Identified 232 student burden reduction policies (1951–2021): 62 national, 170 local (~650,000 words). Notable growth in 2011–2021, with 19 policies in 2021 (~62,000 words). From June–December 2021, local issuances significantly exceeded central issuances. - Governance pattern: Policies follow a “Central leadership + Ministry of Education sovereignty + multi-department coordination” model. In 1981–1990, CPC Central Committee and State Council accounted for 100% of releases; later periods saw increased joint issuances (e.g., Ministry of Finance, HRSS, Market Regulation, Public Security) and MoE-led documents around 50%. - Thematic emphasis via word frequency: High-frequency terms include training (5845), students (3712), institutions (3633), off-campus (3066), school (3041), education (2723), services (2712), after-school (2106), homework (1924), subject (1819), management (1359), charge/fees (979). Parents (856) and society (683) appear far less frequently than students and institutions, indicating limited policy attention to these stakeholders. - Temporal keyword shifts: 1951–1960, “students” >60% of keywords; declined to ~20% by 2011–2021. “Institution” rose markedly since 2011 (66 mentions in 1981–2010; 127 in 2011–2020; 3440 in 2021, ~30% of keywords), surpassing “students” (202 in 2021). “School” stable (25–30% through 1951–2020, <25% in 2021); “teacher” ~10% throughout; “society” increased to >10% in 2011–2020, fell <5% in 2021. “Parents” remained the least frequent (<4% pre-2011; slight increase in 2011–2020, then decline in 2021). - Co-occurrence structures: Across participants (students, institutions, schools, teachers, parents, society), top co-occurring clusters consistently centered on: (1) schools/training/students/education; (2) off-campus/work/services; (3) development/strengthening/after-school. Teachers, parents, society display narrower roles focused on education/teaching. - Parental attitudes: Overall, 84% of parents supported Double Reduction; 16% did not. Significant correlations (p<0.01) between parental support and: parent-rated student burden (r=0.112), homework duration last semester (r=−0.022), sleep duration last semester (r=0.039). Regression shows significant associations for burden, homework (last semester), and sleep (last semester), but not for current semester homework or sleep at 99% CI. - Determinants of support (logistic regression): Female parents were more likely to be non-supportive than males (post-PSM odds reported as ~1.557 times in text; coefficient 1.577 in logit). Urban (non-agricultural) Hukou parents more likely to be non-supportive than rural (post-PSM household type effect ~1.199; 95% CI 1.073–1.340). Higher parental education associated with greater likelihood of non-support. Occupation effects were included but less emphasized. Hosmer–Lemeshow tests indicated good model fit pre- and post-PSM. - Policy-practice gap: Parents and other stakeholders (teachers, society) receive comparatively little attention in policy texts relative to students and institutions, indicating a need to enhance stakeholder engagement.

The study integrates macro policy evolution with micro-level parent perspectives to illuminate how China’s burden reduction policies have transformed alongside socio-economic and ideological shifts. Rapid policy activity, especially since 2011 and with the 2021 Double Reduction, reflects modernization goals for quality education and a governance model that combines central leadership with MoE oversight and multi-agency coordination. Word frequency and co-occurrence analyses reveal policy priorities concentrated on students, institutions, and off-campus training governance, while underemphasizing parents, teachers, and society. This misalignment is consequential, as survey evidence shows parental attitudes are significantly associated with perceived student burden, homework, and sleep. Gender and urban–rural differences indicate that higher expectations and competitive pressures (more acute among female and urban parents) may reduce support for burden reduction policies. The findings suggest that policy success requires addressing multi-level tensions: improving school teaching quality, regulating off-campus training, and actively engaging families and communities. Building a sustainable education ecology necessitates coordinated action across macro (government, economy, culture, environment), meso (family, school, society, after-school), and micro (students, teachers, classroom, management) levels, with greater inclusion of parents to support equitable, inclusive, high-quality, lifelong education.

This study jointly analyzed 232 policy documents (1951–2021) and 23,567 parent responses to map the evolution and on-the-ground perceptions of student burden reduction in China. It identified a governance pattern of central leadership with MoE sovereignty and multi-department coordination, a policy focus on students and institutions with limited attention to parents, and significant links between parental support and student burden indicators. Female and urban parents were less supportive, aligning with higher educational expectations. The results highlight the need to build a collaborative "family-school-society" system that balances in-school quality improvement with out-of-school governance and active parental engagement. Future research should include additional stakeholders (students, teachers, policymakers), cross-country comparisons, and methods that move from correlation to causal inference to better understand drivers of attitudes and outcomes and to refine sustainable, quality-oriented educational reforms aligned with SDG 4.

Key limitations include the reliance on convenience sampling in underdeveloped regions, which may limit generalizability; the absence of teacher survey data, constraining triangulation across major stakeholders; and the predominantly correlational analyses, which limit causal claims. Additionally, word frequency methods, while informative, may underrepresent nuanced policy content and stakeholder roles. Future work should incorporate multi-stakeholder data, probabilistic sampling, and causal designs.