Persistent association between family socioeconomic status and primary school performance in Britain over 95 years

The study asks whether and how the strength of the association between family socioeconomic status (SES) and children’s primary school performance in Britain has changed over the past century. Family SES reflects access to economic, cultural, and social capital and shapes children’s neurocognitive development through multiple pathways. Prior work shows SES-related disparities appear before school entry and often widen through compulsory education, while children’s relative performance remains highly stable from early primary through secondary school. Policies focused on equal opportunities, rising socioeconomic inequality, and increasing parental differential investment (e.g., shadow education) could have altered the SES–achievement link. The authors focus on primary school performance to capture ‘primary effects’ of SES on academic ability, minimize confounding by later educational choices, and leverage the relatively stable structure and aims of primary education. The purpose is to provide a long-run, population-based assessment of temporal stability in the SES–achievement association in Britain.

The paper builds on extensive evidence that SES is linked to academic outcomes through economic, cultural, and social capital and early language and cognitive environments. Prior meta-analyses (White, 1982; Sirin, 2005) reported SES–achievement correlations around r ≈ 0.22–0.28, largely from U.S. samples. Cross-national work (e.g., Chmielewski, 2019) found globally widening SES achievement gaps over ~50 years but stagnation for Britain. Institutional factors like early tracking and market-oriented school policies tend to increase inequality. Broader socioeconomic trends (rising inequality, recessions, COVID-19) may intensify SES effects, while parental investments in shadow education differentially benefit higher-SES children. Conversely, equity-focused policies could reduce SES gradients. Evidence also shows high rank-order stability of school performance across years (correlations ~0.60).

Design: Secondary analysis of 16 British population cohort studies sampling representative populations (city, region, country within the UK, or UK-wide), with births from 1921–2011. Inclusion required at least one valid SES indicator assessed before or concurrent with primary school performance (ages 5–11) and at least one valid school performance measure. Total N = 91,935; cohort Ns ranged from 240 to 14,923. Data sources identified via cohort profiles, UK repositories (e.g., CLOSER), and expert enquiries. Data handling: Each cohort was analyzed separately due to access regulations; harmonization/pooling of raw data was not permitted. Variables were coded so higher scores indicated higher SES/performance and were z-transformed. SES operationalization: Formative indices built from available indicators (mother/father education, occupation, and family income where available), summed and adjusted for number of indicators. School performance: Used earliest post–school entry measures (standardized cognitive/achievement tests, teacher ratings, exam scores, parent reports); where multiple measures existed, built summary indices adjusted for number of measures. Statistical analysis: Within each cohort, computed Pearson correlations between SES and school performance; applied Fisher’s z-transform. Meta-regression using R/metafor adjusted cohort-specific z coefficients for mean-centered confounders: assessment type of performance, number of SES and performance indicators, ages of SES/performance assessment, geographical scope, and percentage missing due to attrition/selective follow-up/linkage. Robustness: (1) Restricted analysis to cohorts with N > 1,000 (k = 11; N = 89,552) born 1946–2011; (2) examined UK-representative cohorts (k = 5; N = 50,306); (3) applied the 90/10 percentile method to estimate performance gaps between lowest and highest SES deciles in large cohorts (N > 1,000), adjusting for the same confounders. Preregistration and data availability: OSF preregistration and materials at https://osf.io/a8fwx/.

- Across 16 cohorts (birth years 1921–2011; N = 91,935), the adjusted average correlation between family SES and primary school performance was r = 0.28 [95% CI 0.22–0.34], a medium effect size. - Adjusted cohort-specific correlations ranged from r ≈ 0.17 [0.10–0.23] to r ≈ 0.37 [0.28–0.46], with no systematic temporal trend; observed differences likely reflect residual cohort-specific confounding rather than time trends. - Robustness (N > 1,000 cohorts; k = 11; N = 89,552): adjusted average r = 0.30 [0.25–0.36]; minimal variation across cohorts (≈0.30–0.31). - UK-representative subset (k = 5; N = 50,306): unadjusted estimate r = 0.33 [0.22–0.43]; adjusted model non-identified due to limited degrees of freedom. - 90/10 percentile method (k = 11; N = 89,552): a minimal apparent reduction over time in the gap between lowest and highest SES deciles, but with large overlapping 95% CIs indicating no reliable trend; estimates more dispersed than correlation-based analyses. - Overall conclusion: The SES–primary school performance association in Britain has been stable in magnitude over 95 years, contradicting expectations of a strong temporal increase.

The findings address the central question by demonstrating temporal stability in the SES–achievement association across nearly a century of British cohorts, despite major societal and policy changes. This corroborates prior British evidence (e.g., Chmielewski’s stagnant gap for the UK) and aligns with meta-analytic estimates of medium-sized SES–achievement correlations. The convergence of correlational meta-regression and the 90/10 method strengthens the inference that no meaningful long-run trend exists. The persistence of the SES gradient implies that policies primarily emphasizing equal opportunities have not translated into reductions in outcome disparities, likely because children arrive at school with differential readiness and resources shaped by SES. The authors argue that achieving more equitable learning outcomes requires reallocating resources to meet differential learning needs, especially early in primary school, to disrupt intergenerational transmission of socioeconomic inequality. Population cohort infrastructures are highlighted as crucial for monitoring change and identifying causal mechanisms over time.

Over 95 years in Britain, the association between family SES and children’s primary school performance has remained consistently of medium magnitude, with no evidence of systematic strengthening or weakening. This stability, despite policy efforts, suggests that focusing on equality of opportunity alone is insufficient. The authors advocate prioritizing equity in learning outcomes by tailoring primary education resources to children’s differential needs to reduce SES-related disparities and weaken intergenerational inequality. Future research should: (1) identify changing mechanisms underlying the SES–achievement link (e.g., language environments, shadow education, school policies); (2) evaluate equity-oriented interventions in early primary settings; (3) enhance temporal coverage with new and continued cohorts; and (4) improve data harmonization and measurement consistency to permit finer-grained causal analyses.

- Heterogeneity across cohorts in measures and ages of assessment (school performance assessed at ages 5–11; performance metrics included tests, grades, teacher ratings; SES indicators varied in type and timing), raising potential residual confounding despite model adjustments. - Some cohorts had relatively small sample sizes, potentially yielding variable power; robustness checks restricted to larger cohorts mitigated but did not eliminate this concern. - Irregular spacing of cohorts with gaps up to 24 years limits detection of finer temporal dynamics. - Inability to harmonize and pool individual-level data across cohorts constrained analytic approaches; analyses relied on cohort-wise estimates and meta-regression. - The adjusted meta-regression model for UK-representative cohorts was non-identified due to limited degrees of freedom. - The study cannot detect changes in underlying mechanisms of SES influence over time; stable association magnitude may mask shifting causal pathways. - Potential biases from attrition, selective follow-up, and data linkage remain despite adjusting for percentage missing data.