Educational inequality remains a central concern in European education systems, particularly in countries characterized by early selection and strong socioeconomic stratification. While family socioeconomic status (SES) is consistently estimated to be a significant correlate of achievement and attainment (see e.g. Liu et al., 2022), recent causal and quasi-experimental evidence suggests that SES has only weak effects on learning rates once children enter school (Song et al., 2025). This discrepancy raises a key theoretical question: if family SES does not substantially shape learning trajectories directly, do SES-related effects manifest through other, indirect channels?

One possible explanation points to the role of educational contexts, especially the socioeconomic composition (SEC) of schools and classrooms. Socioeconomic stratification can influence learning environments through multiple indirect pathways, including peer interactions (Cooc & Kim, 2017), teaching quality (Varga, 2009), teacher expectations (Gentrup et al., 2024), and parental involvement (Barg, 2019).

Since Coleman et al. (1966), SEC has repeatedly been found to predict student achievement over and above individual socioeconomic status, with meta-analytic SEC-achievement correlations far exceeding the SES-achievement correlation levels (Liu et al., 2022). Whether these compositional patterns mark real effects remains debated: observed associations may reflect genuine compositional influence, a statistical artifact of aggregation that disappears once individual SES and prior achievement are adequately controlled, or mean prior ability for which SEC merely serves as a proxy (Marks, 2015; Sciffer et al., 2022).

What unites both sides of this debate is the assumption that classroom composition is adequately summarised by its mean SES. However, this collapses two distinct features,  a class's mean level and the shape of its SES distribution, into a single indicator. Moving beyond these mean-based measures, we introduce SES-distributional representativeness (SES-DR): a distribution-sensitive indicator capturing how closely a classroom's SES distribution resembles that of the wider student population, operationalised as one minus the Kolmogorov-Smirnov distance between the two distributions, residualised on mean class SES. Applied alongside class-mean SES to longitudinal data, it lets us ask whether a classroom's average socioeconomic level, or the breadth of its socioeconomic mix is more important for childern's skill development.

The study addresses two research questions:

1. Do class-mean SES and SES-DR predict reading development between Grades 3 and 8, net of individual SES?

2. Do the associations between class mean SES, SES-DR, and reading growth differ between lower-SES and higher-SES students?

From a theoretical perspective, mean classroom SES and SES-DR can be interpreted as complementary indicators capturing different dimensions of the same contextual effect within the Sen–Bourdieu–Bernstein framework (Hart, 2012; Dimopoulos & Koutsampelas, 2024): they capture the volume and distribution of economic, social, and cultural capital available in the classroom.

The European and international relevance of the study is twofold. First, Hungary represents an extreme but informative case within Europe, as international assessments consistently rank it among the most socioeconomically stratified education systems in the OECD (OECD, 2019). Second, the analytic framework and the SES-DR indicator are portable to other national contexts and international large-scale assessments. By focusing on distribution-sensitive measures of classroom composition, the study contributes to the debate on whether and how socioeconomic mixing can support equity and skill development for low-SES children without high developmental costs for higher-SES children.

The analysis draws on nationwide longitudinal data from the Hungarian Educational Longitudinal Program, following a cohort of students from Grade 3 (fall) to Grade 8 (spring) (N ≈ 6,211 students in 283 classes across 155 schools). Reading literacy was assessed at four time points (Grades 3, 4, 6, and 8) using item-anchored tests aligned with national and international frameworks, with increasing proportions of inferential and evaluative items at higher grades (Hódi et al., 2020, 2025). To place achievement scores on a common developmental scale, we estimated a Rasch model and WLE scores and derived plausible values, standardized to a Grade-3 mean of 500 and a standard deviation of 100 (EAP/PV reliability ≈ .88). Family SES was operationalised as an equal-weight composite of two z-standardised sub-indices: parental education and household material resources. Missing data were handled using multiple imputation with 100 datasets, generated via a random-forest chained-equations algorithm. This approach follows current recommendations for complex longitudinal educational data, allowing for flexible functional forms and interactions, using a large set of auxiliary variables, and reducing bias under missing-at-random assumptions (Wijesuriya et al., 2025). We estimate latent growth-curve structural equation models with empirically determined growth trajectories. Latent intercepts and slopes are regressed on individual SES, mean class SES, SES-DR, and a rich set of student- and class-level controls (e.g., baseline mathematics achievement, self-regulation, class size, gender, and demographic variables). Subgroup-level models are used to assess heterogeneity in the estimated associations across SES-quartiles. Standard fit diagnostics indicate excellent fit (RMSEA = 0.023, CFI = 0.996, SRMR = 0.006).

The analyses show that classroom socioeconomic composition is associated with students' reading development between Grades 3 and 8, net of individual SES, prior achievement, and other controls. Class-mean SES is the dominant predictor: it is positively related to reading growth across the entire SES distribution (b ≈ 0.14; p < .001), and this association strengthens rather than attenuates once class-mean prior achievement is controlled for. Beyond mean SES, SES-DR shows a positive but non-significant association with reading growth in the pooled sample. Heterogeneity analyses confirm this pattern: class-mean SES predicts faster growth across all SES quartiles, with the largest estimate among lower-middle-SES students, while SES-DR is never negative in any subgroup. Among classes with similar mean SES levels, socioeconomic mixing therefore does not slow the reading development of higher-SES students, while placement in higher-SES classrooms supports growth across the spectrum. This means that the level of the classroom, and not the breadth of its SES distribution, is what matters.

Barg, K. (2019). Why are middle-class parents more involved in school than working-class parents? Research in Social Stratification and Mobility, 59, 14–24. https://doi.org/10.1016/j.rssm.2018.12.002 Coleman, J. S., Campbell, E. Q., Hobson, C. J., McPartland, J., Mood, A. M., Weinfeld, F. D., & York, R. L. (1966). Equality of educational opportunity. U.S. Government Printing Office. Cooc, N., & Kim, J. S. (2017). Peer influence on children’s reading skills: A social network analysis of elementary school classrooms. Journal of Educational Psychology, 109 (5), 727–740. https://doi.org/10.1037/edu0000166 Dimopoulos, K., & Koutsampelas, C. (2024). Towards a new approach for understanding educational inequality: The Sen-Bourdieu-Bernstein theoretical framework. International Journal of Educational Research, 125, 102347. https://doi.org/10.1016/j.ijer.2024.102347 Gentrup, S., Olczyk, M., & Lorenz, G. (2024). Teacher stereotypes and teacher expectations at the intersection of student gender and socioeconomic status. Zeitschrift für Entwicklungspsychologie und Pädagogische Psychologie, 56 (1–2), 87–102. https://doi.org/10.1026/0049-8637/a000291 Hart, C. S. (2012). Aspirations, education and social justice: Applying Sen and Bourdieu. Bloomsbury Publishing. Hódi, Á., Tóth, E., & Nikolov, M. (2025). Interactions among text formats and reading processes in young learners’ L1 reading literacy. Reading Research Quarterly, 60 (1). https://doi.org/10.1002/rrq.595 Liu, J., Peng, P., Zhao, B., & Luo, L. (2022). Socioeconomic status and academic achievement in primary and secondary education: A metaanalytic review. Educational Psychology Review, 34 (4), 2867–2896. https://doi.org/10.1007/s10648-022-09689-y Marks, G. N. (2015). Are school-SES effects statistical artefacts? Evidence from longitudinal population data. Oxford Review of Education, 41 (1), 122–144. https://doi.org/10.1080/03054985.2015.100661 OECD. (2019). Pisa 2018 results (volume II): Where all students can succeed. https://doi.org/10.1787/b5fd1b8f-en Sciffer, M. G., Perry, L. B., & McConney, A. (2022). The substantiveness of socioeconomic school compositional effects in Australia: Measurement error and the relationship with academic composition. Large-Scale Assessments in Education, 10, 21. https://doi.org/10.1186/s40536-022-00142-8 Song, Q., Liu, Y., & Tan, C. Y. (2025). Effects of family socioeconomic status on educational outcomes in primary and secondary education: A systematic review of the causal evidence. Educational Psychology Review, 37 (2). https://doi.org/10.1007/s10648-025-10004-8 Varga, J. (2009). A tanárok elosztása a különböző szociokulturális hátterű tanulókat tanító iskolák között [Distribution of teachers among schools teaching students from different sociocultural backgrounds]. In K. Fazekas (Ed.). Oktatás és foglalkoztatás [Education and employment]. HAS Institute of Economics. Wijesuriya, R., Moreno-Betancur, M., Carlin, J. B., White, I. R., Quartagno, M., & Lee, K. J. (2025). Multiple imputation for longitudinal data: A tutorial. Statistics in Medicine, 44 (3–4). https://doi.org/10.1002/sim.10274