A 2025 administrative-data study of all children aged 8 to 15 in the Madrid Community (2016–2019) found that the effect of the parental environment on test scores evolves differently by school subject: it declines with age in mathematics, follows a bell curve in literature, and increases steadily in English as a foreign language — a Matthew effect (a tendency for initially advantaged children to pull further ahead) present only in foreign-language learning. The study's instrument, the 1960 gender gap in tertiary education in the parents' country of origin, rests on an exclusion restriction that cannot be tested directly, and its authors explicitly flag what remains unresolved: validating or replacing the instrument, disentangling the reverse-causal reading of early-grade homework time, establishing external validity beyond one region and period, decomposing the parental effect into genetic and environmental transmission channels, and identifying the mechanisms behind the subject-differentiated age trajectories. This problem record states that surviving open core: establishing, with credible causal identification and beyond a single region, whether and how the parental-environment effect on achievement evolves with child age across subjects, and which transmission channels account for the observed subject differentiation.

Background

The term "Matthew effect" was introduced by Robert K. Merton (1968) to describe how initial advantage compounds into further advantage; in education research it denotes the hypothesis that achievement gaps linked to family background widen as children grow older, so that early differences in parental environment amplify over schooling. Two strands of literature motivate the question. In the economics of skill formation, Cunha and Heckman (2007) model cognitive and non-cognitive skill formation as a dynamic technology in which skills produced at one stage raise the productivity of later investment (self-productivity and dynamic complementarity), which mechanically predicts widening gaps — a Matthew effect — unless compensating forces intervene. In empirical education research, the Matthew-effect hypothesis has been tested almost exclusively for reading: a review of 25 years of longitudinal reading research (Pfost et al. 2014) concluded that evidence for widening gaps is mixed, and Protopapas et al. (2011) found no genuine Matthew effect in reading comprehension. Whether the parental-environment effect grows, shrinks, or is stable with age — and whether the answer differs across core school subjects on a unified database — was untested before Aymeric, Lavaine and Magdalou (2025).

That study uses administrative records covering all children aged 8 to 15 in the Madrid Community from 2016 to 2019, with scores in mathematics, literature (Spanish), and English as a foreign language at three grade levels (Grades 3, 6, and 10). "Parental environment" is operationalized as parents' highest level of education, classified in three ISCED categories (the UNESCO International Standard Classification of Education levels 0–2, 3–5, and 6–8), plus measures of parental investment such as how often parents talk to their child about school. Because parental education is not randomly assigned — it correlates with genetic endowments, income, and other family characteristics — the authors complement ordinary least squares with an instrumental-variable (IV) strategy: two-stage least squares in which parents' education is instrumented by the gender gap in tertiary education in 1960 in the parents' country of origin (GGGG), constructed from Barro–Lee educational attainment data (a standard cross-country dataset of educational attainment by sex and age group). The identifying assumption (an exclusion restriction) is that the 1960 country-of-origin gender gap affects the child's scores only through the parents' attained education. The estimated effect of parental education is positive and significant at all ages in all subjects (the persistent effect), but its evolution with age differs sharply by subject: declining in mathematics, bell-shaped in literature, and significantly increasing in English — so a Matthew effect in Merton's sense is supported only for foreign-language learning, consistent with the social, interaction-heavy nature of second-language acquisition.

For a reader from a neighboring field, the relevant conventions are: an exclusion restriction is inherently untestable and is defended indirectly (here, by arguing that the 1960 gender gap is only weakly correlated with country wealth and is driven by social norms, citing Huber and Paule-Paludkiewicz 2024); a first-stage F-statistic far above conventional thresholds rules out weak instruments but says nothing about exclusion; and single-region administrative data raise external-validity questions that only replication elsewhere can settle. The broader causal literature on parental schooling (surveyed by Holmlund, Lindahl and Plug 2011, who compare IV, adoption, and twin designs) shows that the gene–environment decomposition of parental-education effects remains contested: adoption-based evidence attributes a substantial share of the parent–child education correlation to genetic transmission, and the Madrid study cannot separate genetic from environmental channels because it observes neither adoption status nor genetic data.

Problem Statement

A 2025 administrative-data study (Aymeric, Lavaine and Magdalou, arXiv:2510.18481) documents, on records covering all children aged 8 to 15 in the Madrid Community (2016–2019, tested in Grades 3, 6, and 10), that the estimated effect of parents' education on student scores declines with age in mathematics, is bell-shaped (first increasing, then decreasing) in literature, and increases steadily in English as a foreign language, with instrumental-variable estimates using the 1960 country-of-origin tertiary-education gender gap reinforcing these patterns. The open problem is the full causal and mechanistic establishment of this subject-dependent evolution of the parental-environment effect. Its open core comprises: (1) whether these age trajectories are credibly causal — in particular whether the exclusion restriction of the gender-gap instrument (that the 1960 country-of-origin gender gap affects child outcomes only through parents' attained education) survives scrutiny against alternative channels such as country-of-origin wealth, social norms, or education-selective migration, or whether the subject-differentiated trajectories survive under other credible identification strategies; (2) whether the trajectories generalize beyond the Madrid Community and the 2016–2019 period; and (3) which transmission channels — genetic versus environmental (parental investment, attitudes, assortative mating), and mechanisms such as non-cognitive-skill productivity, school selection, or peer effects — account for why the parental effect attenuates with age in mathematics, first rises and then falls in literature, but amplifies in foreign-language learning, resolving in particular the reverse-causal ambiguity of the early-grade homework–score association (compensatory homework by struggling pupils). These three components are jointly posed: a full account requires credible causal identification of the subject-dependent trajectories, their external validity, and a mechanism account of the subject differentiation. Whether targeted remediation or early foreign-language programs reduce the resulting subject-specific inequalities in opportunity is an explicitly downstream follow-up application, not part of this core. The common objective uniting these components is a causally identified, externally valid, and mechanistically explained account of how the parental environment's impact on achievement evolves as children grow, across academic subjects.

References

  1. Gaëlle Aymeric, Emmanuelle Lavaine, Brice Magdalou (2025). Parental environment and student achievement: Does a Matthew effect exist? arXiv:2510.18481. https://arxiv.org/abs/2510.18481
  2. Maximilian Pfost, John Hattie, Tobias Dörfler, Cordula Artelt (2014). Individual Differences in Reading Development: A Review of 25 Years of Empirical Research on Matthew Effects in Reading. Review of Educational Research 84(2):203–244, DOI 10.3102/0034654313509492. https://doi.org/10.3102/0034654313509492
  3. Athanassios Protopapas, Georgios D. Sideridis, Angeliki Mouzaki, Panagiotis G. Simos (2011). Matthew Effects in Reading Comprehension: Myth or Reality? Journal of Learning Disabilities 44(5):402–420, DOI 10.1177/0022219411417568. https://doi.org/10.1177/0022219411417568
  4. Helena Holmlund, Mikael Lindahl, Erik Plug (2011). The Causal Effect of Parents' Schooling on Children's Schooling: A Comparison of Estimation Methods. Journal of Economic Literature 49(3):615–651, DOI 10.1257/jel.49.3.615. https://doi.org/10.1257/jel.49.3.615
  5. Flavio Cunha, James Heckman (2007). The Technology of Skill Formation. American Economic Review 97(2):31–47, DOI 10.1257/aer.97.2.31. https://doi.org/10.1257/aer.97.2.31
  6. Stefanie J. Huber, Hannah Paule-Paludkiewicz (2024). Gender norms and the gender gap in higher education. Labour Economics 87:102491, DOI 10.1016/j.labeco.2023.102491. https://doi.org/10.1016/j.labeco.2023.102491
  7. Robert K. Merton (1968). The Matthew Effect in Science. Science 159(3810):56–63, DOI 10.1126/science.159.3810.56. https://doi.org/10.1126/science.159.3810.56