The effect of intergenerational mobility on family education investment: evidence from China

China’s rapid growth since 1978 has coincided with widening income inequality, shifting families’ educational spending toward extracurricular activities. While family background factors influencing education investment are well studied at the micro level, macro-level determinants—particularly intergenerational mobility (IGM) as a proxy for equality of opportunity—are less explored. IGM is considered a key indicator of societal equality of opportunity and is linked to inequality and subjective well-being. This paper asks: (1) Does opportunity inequality (via IGM) affect family education investment (FEI)? (2) Through which mechanisms—greater incentives versus reduced anxiety/status-seeking—does IGM operate? (3) Is the effect heterogeneous across SES groups given differences in borrowing constraints and incentives? Using city-level IGM from the 2015 1% Population Sample Survey and FEI and covariates from CFPS 2018, the study tests whether higher IGM affects FEI, explores mechanisms, and examines heterogeneity by SES and urban–rural status.

Prior work suggests higher equality of opportunity could motivate greater education investment because individuals’ achievements depend more on effort (Browman et al., Mae, Wen & Witteveen). Evidence on income inequality’s effect on FEI is mixed: some find increased educational spending due to status-seeking under greater inequality; others find credit constraints depress optimal investment among poorer families. Recent studies link competitive cultures to parental educational anxiety, which can raise extracurricular investment but also exacerbate social inequality. Given IGM is a measure of equality of opportunity and is related to inequality (Great Gatsby Curve), it may affect FEI via incentive, anxiety, and status-seeking channels. Few studies directly test IGM’s impact on FEI; Wen and Witteveen (2021) considered perceived mobility but did not address status/anxiety mechanisms or endogeneity. This study aims to fill these gaps with objective IGM and FEI measures and causal identification strategies.

IGM measurement: Intergenerational educational mobility is measured using intergenerational rank correlation at the city level following Dahl and DeLeire (2008). For each city c, regress child’s education rank on parent’s education rank: Rank_child_ic = α_c + β_c Rank_parent_ic + ε_ic. IGM_c = 1 − β_c, with higher values indicating greater mobility. Education ranks are constructed within city-cohorts based on years of schooling, leveraging the discreteness of educational attainment and the availability of parent–child educational links in the 2015 1% Population Sample Survey. Floating populations, individuals under 16, and those still in school were excluded; only heads of household, their parents, and children were kept to match pairs. City-specific β_c are estimated, then IGM computed. Data: (1) China 1% Population Sample Survey 2015 to compute city-level IGM; (2) CFPS 2018 for FEI and covariates; (3) city-level GDP and public education finance from yearbooks and city statistics; city-level Gini calculated from CFPS for robustness. Dependent variable (FEI): Extracurricular/home tutoring expenditure for children aged 0–16 over the past 12 months from CFPS 2018 child proxy questionnaire. FEI transformed as ln(FEI + 1) to address zeros and heteroscedasticity. A binary indicator (dFEI) equals 1 if FEI>0, else 0, is used in robustness checks. Key independent variable: City-level IGM from 2015 survey. An improved IGM (ImIGM) controls for lifecycle bias by including parents’ and children’s ages and squared terms in rank estimation. Controls: Child-level (age, gender, learning stage, urban hukou, school type key/model, mother’s education); family-level (annual income, number of children, household size, head’s gender/age, parental education expectations, number of books); city-level (financial allocation per compulsory student). Province fixed effects; standard errors clustered at the city level. Empirical models: Baseline OLS and Tobit (for censoring at zero): ln(FEI_ijcp + 1) = β0 + β1 IGM_c + βx X_i + φ X_j + γ X_p + v_p + ε_ijcp. Robustness includes: (a) replacing dependent variable with dFEI via OLS/Probit; (b) replacing independent variable with ImIGM; (c) adding city GDP and Gini; (d) assessing bias from unobservables using the Altonji–Elder–Taber/Bellows–Miguel approach with ratios comparing restricted/full control sets; (e) Instrumental variables (2SLS and IV Tobit) using as instrument the mean IGM of other cities within the same province (leave-one-city-out provincial average), justified by inter-city competition for educational resources affecting IGM but not directly city FEI. Heterogeneity: 2SLS by subsamples: urban vs rural; high vs low cultural capital (books above/below median); high vs low economic capital (income above/below median). Mechanisms: 2SLS/IV Probit using same IV. Incentive: parental agreement that “hard work is rewarded” (1–5) for mothers (DIM) and fathers (DIF). Excessive educational anxiety (EEA): binary indicator based on whether parents paid extra fees for school choice. Status-seeking: difference between parents’ actual income rank (city deciles 1–10) and self-rated local income status (1–5), scaled by self-rating; higher values indicate stronger status-seeking. Outcomes regressed on IGM controlling for the same covariates and province fixed effects.

- Baseline effect: Higher intergenerational mobility significantly reduces family education investment. OLS with full controls: coefficient on IGM = −2.575 (SE 1.041); a 0.1 increase in IGM reduces FEI by about 25.75%. Tobit results are consistent (e.g., −2.795, SE 1.122). - Robustness to alternative measures: Using dFEI (participation) via OLS/Probit shows negative, significant effects of IGM (e.g., −0.295 OLS; −0.413 Probit). Using improved IGM (ImIGM) yields similar negative coefficients in OLS/Tobit. - Controlling omitted variables: Adding city GDP and/or Gini leaves the IGM effect negative and significant or marginally significant across specifications; GDP positively relates to FEI; higher Gini tends to associate with higher FEI. - Unobservables assessment: Ratios comparing shifts from restricted to full controls indicate that unobservables would need to be multiple times as correlated as observables to explain away the effect (e.g., ratios often >4, up to ~30), supporting robustness. - Instrumental variables: 2SLS first-stage F-statistics far exceed 10 (e.g., 222.7; IV Tobit first-stage F ≈ 702.2). 2SLS coefficient on IGM ≈ −2.563 (SE 0.901), IV Tobit ≈ −2.842 (SE 1.225), confirming a significant negative causal effect. - Heterogeneity: Significant negative effects are concentrated in urban families and in families with high cultural capital (books) and high economic capital (income). Effects are not significant for rural, low-cultural-capital, or low-income families, consistent with credit constraints and stronger mobility incentives among lower-SES households. - Mechanisms: IGM increases fathers’ incentive (DIF; coefficient ~0.855, p<0.1), but not mothers’. IGM reduces excessive educational anxiety (EEA) in IV Probit (negative and significant). IGM reduces fathers’ status-seeking (SSF; coefficient ~−1.689, p<0.1), with no significant effect for mothers. Net effect on FEI is dominated by reductions in excessive anxiety and status-seeking among higher-SES families. - Descriptive range of IGM: City IGM ranges from 0.321 to 0.826 (difference 0.505); sample mean ~0.615 (SD 0.114). Mean ln(FEI) ≈ 6.783 in the analytic sample.

The study demonstrates that greater intergenerational mobility—indicative of higher equality of opportunity—reduces family spending on extracurricular tutoring. This answers the primary research question by showing a negative causal effect of IGM on FEI. The mechanism analysis clarifies that, while IGM may enhance beliefs in effort-based success (particularly among fathers), the dominant pathways reduce FEI by alleviating excessive parental educational anxiety and status-seeking behavior. Heterogeneous effects indicate that high-SES and urban families, who have fewer credit constraints and engage more in competitive status-driven investments, cut back more when opportunities are more equal. Low-SES families, constrained by finances and with higher incentives from mobility prospects, show no statistically significant response. These findings suggest that enhancing equality of opportunity can temper positional competition and reduce overinvestment in shadow education, aligning with the goals of China’s “Double Reduction” policy. The results contribute to the literature by identifying new channels—anxiety and status-seeking—through which macro-level opportunity structures shape micro-level parental investment decisions.

This paper provides one of the first causal estimates of the impact of intergenerational mobility on family education investment in China. Using city-level educational rank-based IGM, CFPS measures of extracurricular tutoring expenditure, and multiple identification strategies (OLS, Tobit, 2SLS/IV Tobit), it finds that higher IGM significantly reduces FEI, particularly among urban and higher-SES households. Mechanism analysis indicates reductions in excessive educational anxiety and status-seeking, with some increase in incentive beliefs among fathers. Policy implications include promoting equality of opportunity, increasing public education spending, more efficient allocation of educational resources, and reducing parental anxiety to support the effective implementation of the “Double Reduction” policy. Future research should explore additional mechanisms (e.g., returns to education, income dynamics), utilize panel data with fixed effects to further address endogeneity, and investigate effective policies to enhance intergenerational mobility.

- Potential omitted mechanisms: The analysis focuses on incentives, educational anxiety, and status-seeking; other channels such as changes in the return to education or family income dynamics may also mediate effects. - Data design: The study relies on cross-sectional data for both IGM (2015) and FEI (2018); lack of panel identification limits the ability to control for time-invariant unobserved heterogeneity at the household or city level. - Measurement constraints: While rank-based educational IGM mitigates some biases, lifecycle and measurement errors may persist despite improvements. FEI measured via extracurricular spending may miss other forms of educational investment. - Generalizability: Results pertain to China’s institutional context (urban–rural dual structure, shadow education market, “Double Reduction” policy) and may not directly generalize to other countries without similar structures.