Are migrants a threat? Migrant children and human capital investments among local households in urban China

The study investigates whether and how the presence of internal migrant children in urban Chinese classrooms influences human capital investment decisions of local (native) households. While prior work has focused primarily on the effects of migrant peers on natives’ academic outcomes, little is known about how native families adjust educational spending and time investments in response to migrant classmates. This gap matters because behavioral spillovers may represent important social externalities of migration that are overlooked when only test scores are studied. China provides a salient setting given large-scale internal migration, the hukou system’s role in allocating public services (including education), and the prevalence of intensive shadow education. The authors pose two questions: (1) Do migrant children generate spillover effects on local households’ human capital investments? (2) If so, through what mechanisms do these reactions arise? Exploiting random within-school classroom assignment, they aim to identify causal effects and illuminate whether native parents’ responses are grounded in actual changes in classroom environments and achievement or in biased beliefs and concerns.

The paper reviews evidence on educational spillovers from migrant peers, noting mixed findings in international contexts: some studies document negative impacts on natives’ outcomes (e.g., in the U.S., parts of Europe, Israel, Denmark), while others find negligible or even positive effects (e.g., UK, Netherlands, Australia, Norway). For internal migration, the evidence base is smaller and similarly mixed. In China, studies using random class assignment report heterogeneous results regarding natives’ test scores in response to internal migrant peers. Potential channels include resource strain in schools receiving migrants, teachers adapting pedagogy for lower-prepared students, and peer composition effects on learning attitudes. The authors also connect to literature on household human capital investment strategies (reinforcement vs compensatory), noting SES-related heterogeneity, and to studies on native flight and parental responses to perceived declines in school quality when migrant shares rise. They discuss an alternative ‘competition effect’ hypothesis: migrant inflows potentially heighten perceived labor market competition, prompting additional investments by natives. Finally, the paper provides background on China’s hukou system, access barriers for migrant children to public schools, and prior findings that between-school differences largely account for migrant-local achievement gaps, underscoring the policy importance of school access quality for migrants.

Data come from the China Educational Panel Survey (CEPS), a nationally representative longitudinal survey of middle school students initiated in 2014, with detailed information on students, parents, teachers, and school principals. The analysis focuses on public middle schools in urban areas and exploits within-school random assignment of students to classes. Schools are retained if principals report random assignment and fixed class configurations; to guard against misreporting, the authors further restrict to schools where all teachers in the same grade confirm that assignment is not based on test scores. After restrictions, the analysis spans 41 schools and 152 classes. The main estimation sample for local students includes 3,515 observations (with migrant peers present in the same classes). The study measures household human capital investments via parental reports: total education expenditure (semester), in-school expenditure (tuition, textbooks, uniforms, accommodation, insurance), out-of-school expenditure (extracurricular classes, private tutoring, hobby training), and parental time investment. The key independent variable is the class-level proportion of migrant children. Identification relies on random assignment within grade-school cells, verified by balance tests at student, family, and class levels (Pearson χ² and OLS checks), and grade-by-school fixed effects control for group-level confounders. The baseline model is an OLS regression: H_ij = α + β propMC_g + X_ij γ + D_p + ε_ij, where H_ij denotes household investments for local student i in class j; X_ij includes predetermined student, family, and class characteristics; D_p are grade-by-school fixed effects; and standard errors are clustered at the class level. Robustness checks include alternative definitions of migrant children (rural vs urban origin; living with at least one parent), subsamples (schools in city centers), additional contextual controls, conditioning on parental knowledge of children’s friends, inclusion of migrant students in the sample, and Oster (2019) sensitivity analyses. Mechanisms are probed via: (a) parental beliefs regarding their child’s achievement and school environment; (b) alternative peer group definitions based on baseline academic rank proximity to test for ‘competition effects’; (c) teachers’ reports on pedagogy and class learning environments; and (d) students’ reported school life quality, emotional well-being, and measured academic test scores. Outcomes are standardized where noted; fixed effects and controls are consistently applied.

• A higher share of migrant children in a class causally increases local households’ education spending, with especially strong effects on out-of-school (shadow) education. The authors quantify that a one-percentage-point increase in migrant share raises total education expenditure by about 2.3% and out-of-school expenditure by about 4.8% for local students. Baseline OLS estimates using the random-assignment subsample corroborate significant positive effects on total and out-of-school expenditures; parental time investment does not significantly change.
• Robustness: Results hold across alternative measures of tutoring (e.g., more time spent in tutoring on weekdays and weekends, more tutoring subjects, higher likelihood of private tutoring), alternative migrant definitions (rural and urban migrants both drive increases, with larger effects for rural migrants), subsamples (schools in city centers), and models with additional contextual controls. Oster bounds suggest limited scope for omitted-variable bias to overturn conclusions.
• Heterogeneity: Effects are larger for boys, for ninth graders (facing high school entrance exam pressure), and for higher-SES families (measured by higher parental education and better reported family economic status).
• Mechanisms: Parental belief channels dominate. As migrant share rises, local parents increasingly overestimate their own child’s academic standing and express concerns that migrant presence harms educational quality and school atmosphere; they also perceive teachers as less responsible/patient toward their child. These beliefs correspond with increased private tutoring expenditures. By contrast, there is little to no evidence that migrant peers significantly deteriorate teachers’ reported classroom learning environments or pedagogical quality.
• Competition effect not supported: When redefining peers by baseline academic rank proximity, effects are smaller among the nearest academic competitors and become larger as the peer set widens, with stronger impacts tied to migrant peers with lower achievement rather than higher-achieving competitors. This pattern is inconsistent with a primary competition mechanism.
• Student outcomes: No significant effects of migrant peer share on local students’ school-life quality, emotional well-being, or test scores (Chinese, math, English, average academic, cognitive), both contemporaneously and in one-year follow-up analyses; controlling for household investments does not alter these null result patterns.

The findings directly address the research questions by showing that migrant peer composition elevates native households’ financial investments in children’s education, particularly shadow education. However, these behavioral responses are not justified by measurable declines in classroom environments, student well-being, or academic achievement. Instead, they appear driven by biased parental beliefs—overestimation of their child’s academic standing and fears that migrant peers harm educational quality—which trigger precautionary or compensatory spending, especially among high-SES families and in high-stakes grades. The results nuance the debate on educational inclusion of internal migrant children in China, indicating that natives’ increased spending is not a response to actual learning harms but to perceptions and concerns. This has implications for social integration and education policy: misperceptions can fuel inefficient household spending without improving outcomes, suggesting a role for information and communication strategies, alongside structural policies ensuring migrant access to quality public schooling.

This study demonstrates that the presence of internal migrant classmates in urban Chinese middle schools causally increases local households’ educational expenditures—most notably on out-of-school tutoring—without discernible negative effects on local students’ classroom environments, emotional well-being, or academic performance. The evidence points to parental belief-based mechanisms rather than actual educational harms or competition effects: local parents overestimate their child’s relative performance and worry about migrant-related threats to educational quality, leading to higher spending. Contributions include: (1) shifting focus from achievement to behavioral spillovers in receiving communities; (2) clarifying mechanisms behind household investment responses; and (3) informing policy on migrant educational inclusion and parental information. Future research should investigate heterogeneity by migrants’ duration of residence, broader forms of time investments (including other caregivers), other educational stages beyond middle school, and external validity across different migration contexts and education market competitiveness levels.

• Migration exposure heterogeneity: The study cannot examine differences by migrants’ duration of local residence due to data limitations.
• Time investment measures: Lack of information on broader household time inputs, including contributions from additional caregivers (e.g., grandparents), limits analysis of non-financial investments.
• Sample scope: The sample focuses on middle school students; findings may not generalize to primary or high school contexts or to other age groups.
• Context specificity: Results pertain to internal migration within China’s hukou and education systems; external validity to international migration or education systems with different competitiveness may be limited.
• Some reported sample counts across sections/tables indicate minor inconsistencies; however, the identification relies on within-school random assignment and is supported by extensive balance and robustness checks.