Key result — Intergenerational vs. multigenerational mobility
Parent–child (G1–G2) correlation in years of education ≈ 0.516 (average) and G2–G3 ≈ 0.638, indicating relatively low parent–child mobility in Indonesia .
Predicted (naïve) G1–G3 = product of two-generation estimates; observed three-generation correlations (~0.25–0.31) are slightly below or similar to naïve predictions, so Indonesia shows faster multigenerational decay than many high-income regions .
Multigenerational regression: baseline estimates
Specification
Parental Education (βp)
Grandparental Education (βgp)
Baseline OLS
0.502***
-0.0426* (negative)
Notes: Dependent = child years of education; standard errors clustered by family; results robust to ethnicity and district FE and alternative specs. The negative βgp (conditional on parents) is statistically significant in baseline specifications and in LPM for high-school completion .
Theoretical mechanisms and predictions
Becker–Tomes variant with a direct parental income effect (γ>0) predicts βgp can be negative because direct income effects can mask transferable endowments; by contrast, latent-factor models (measurement error / latent endowments) predict positive βgp; the original Becker–Tomes (stochastic) nests both outcomes, so sign of βgp depends on relative sizes of γ, λ, ρ .
Comparative implication: stronger direct income/channel (e.g., due to binding credit constraints) → larger parent–child correlation but more negative βgp (faster multigenerational decay) .
Evidence on financial constraints — crisis DID (1997–98)
Identification: kabupaten-level proportionate change in DMSP-OLS nighttime lights 1996→1998 as a measure of crisis exposure, cohorts split by age when crisis hit (Born ≤1978 vs. Born ≥1979), DiD interacted with parental and grandparent education .
Main empirical pattern: β on (GP edu × crisis × treated cohort) is negative for boys who were in lower/upper secondary when the crisis hit (statistically significant), consistent with financial constraints increasing the magnitude of negative grandparent coefficient for affected cohorts; effects are weaker or of opposite sign for girls .
Assortative mating / marital customs — evidence and implications
IFLS directly asks who selected the spouse; ~28% of women and ~19% of men report family selected spouse, with strong province/ethnicity variation .
Empirical facts: family-based selection ⇢ weaker spousal education correlation but stronger correlation between a spouse and parents-in-law; interacting FamilyChoice with regressors shows that grandparental education becomes more predictive when parents were matched by grandparents (FamilyChoice), i.e., family-based matching amplifies multigenerational persistence .
Theoretical mapping: family-based assortative matching can reduce observed parent–child β−1 but increase β−2, thus raising βgp relative to direct assortative mating (model derivations in Appendix) .
Data & robustness
Data: IFLS pooled waves (1993,1997,2000,2007,2014), restricted sample with complete three-generation education (N=8,278); years-of-education constructed by mapping highest level to typical years; validated against census aggregates .
Robustness: results stable across FE (ethnicity, kabupaten), nonlinear specifications, lineage-specific regressions, and placebo checks; crisis measure validated vs. province/GDP and Bartik-style shocks (Appendix Figures/Tables) .
Policy-relevant takeaways
High parent–child persistence in developing countries does not mechanically imply low long-run mobility — the dynamics hinge on mechanisms like credit constraints and marital institutions, which have distinct multigenerational signatures .
Interventions altering credit access for families (scholarships, borrowing facilities) or shifts in marital customs may change not only short-run upward mobility but also long-run persistence profiles .
思维导图
深度阅读
以下为对“Educational Mobility Across Multiple Generations in Indonesia”研究报告的逐节、图表与方法的详尽解构与评价,结构按照报告自身章节展开,并在每处直接依据原文所得结论/推断处以页码溯源标注(每句末标注来源页码)。本文力求逐点剖析作者的论点、数据、模型、表格与图形,并对其假设、计量策略、识别与稳健性进行审慎评估。
一、元数据与总体概览(题名、作者、主题与核心结论)
报告标题、作者与出版信息:题为“Educational Mobility Across Multiple Generations in Indonesia”,作者为 Sarah Cattan、Antonio Dalla‑Zuanna、Jan Stuhler、Po Yin Wong,发布日期为 2026‑04‑23,研究主题聚焦印尼三代教育流动性与多代传递机制。
仍需进一步工作:建议未来研究尝试(i)结构估计以直接估计 γ、λ、ρ 的规模以更明确量化机制贡献;(ii)利用更多外生冲击或政策(如学校建设、信贷扩张)做更强的因果识别;(iii)更深入考察性别差异的机制(偏好 vs 回报差异 vs 文化规范)。这些建议均基于报告中提出的理论路径与未完全解析的实证异质性。