Single-Sex Schooling and Academic Attainment at School and Through the Lifecourse

A nationally representative birth cohort tracked for 46 years, with standardized test scores collected before secondary school began and exam results from school records, is a strong observational foundation. The historical context — strict catchment-area rules, no school league tables, limited housing mobility — creates a quasi-natural-experiment setting better than most studies of school type.

Three problems limit it. First, the evidence that students at single-sex and co-educational schools were similar within each school sector is referenced but never shown. The reader is told "little difference was found" without being able to verify what that means. Second, most schools are probably represented by a single student, so school-level characteristics — teacher quality, disciplinary climate, resources — cannot be controlled or described. Third, the "treatment" is a bundle: a girls-only school in the 1970s differed from a co-educational school in many ways beyond the absence of boys. The basic-skills measures at age 42 are too crude to detect any effect, making those null findings uninformative.

The regression framework is well-chosen: binary logistic regression for binary outcomes, partial proportional odds models for ordered counts where the standard proportional-odds assumption was violated. The large sample (12,320) is adequate for the main findings. Odds ratios are reported consistently, and interactions are tested and reported transparently.

Several problems weaken the picture. No confidence intervals accompany any estimate, so the reader knows whether results cleared a threshold but not how precise they are. No power analysis supports the null findings — the paper concludes single-sex schooling had no effect on boys' attainment, on post-16 outcomes, or on basic skills, without estimating what size of effect the study could have detected. No correction for multiple comparisons is applied despite many outcome models, and one finding (men's likelihood of female-typed qualifications) moves in and out of conventional thresholds depending on model specification. The unexplained "direct effect" of single-sex schooling on women's qualification field at age 33 is presented as evidence of a lasting school effect without considering that an unmeasured confound could produce the same pattern.

Measurement quality is high where it matters most: exam results at ages 16 and 18 come from school records, and prior-attainment controls use standardized tests at ages 7 and 11. The age-33 classification of qualification fields into male-dominated, female-dominated, and integrated categories is coarser but serviceable.

The distance between what is measured and what is claimed widens for the causal conclusions. The finding that students at single-sex schools achieved more in gender-atypical subjects is well-documented as an association. But saying co-education "exacerbated" gendered patterns implies a causal direction the design cannot establish — single-sex schools may have reduced differentiation, co-ed schools may have increased it, or an unmeasured confound may explain the difference. The claim that single-sex environments "reduce the tendency of students to behave according to gender-typical stereotypes" is the widest leap: no measure of stereotype endorsement or stereotype-driven decision-making exists in the dataset, and the paper's mediation analysis found that self-concept at 16 did not explain the age-33 effect. If reduced stereotyping were the mechanism operating through self-concept, that mediation should have appeared.

The paper reasons carefully through its descriptive findings and is honest about threshold sensitivity, null findings, and the limits of extrapolating from one era to another. It distinguishes raw differences from regression-adjusted ones, noting that the apparent advantage of single-sex schools in raw exam scores is "extremely misleading" given the concentration of single-sex schools in selective sectors.

Two reasoning problems stand out. First, the entire study depends on the claim that within-sector selection was innocuous, but the supporting evidence is withheld. Second, the unexplained age-33 effect — women from single-sex schools were more likely to hold qualifications in male-dominated fields, and this was not explained by exam results or self-concept at 16 — is interpreted as a "direct effect" of schooling without considering the obvious alternative: an unmeasured confound (family gender attitudes) drove both school selection and later career-field choice. The conclusions also use causal language ("exacerbated," "reduce the tendency") that the observational design does not support, though the body text is more careful than the summary statements.

The study uses a publicly accessible (though not open-access) national dataset, and the outcome measures are objective where they matter most. But multiple standard elements are missing: no pre-registration, no analysis code, no supplementary materials, no power analysis for null findings, no conflict-of-interest or funding disclosure, no ethics statement. Most consequentially, the "preliminary analyses" supporting the paper's central methodological argument — that within each school sector, students at single-sex and co-ed schools were comparable — are referenced but never shown. The reader must accept this foundational claim on trust.

The literature review is balanced for a politically charged topic, representing feminist arguments for and against single-sex schooling, conservative positions, and the essentialist "different brains" perspective. Prior work on the same dataset (Steedman 1983) is honestly acknowledged and differentiated from the current study. The one citation verifiable from source material (Lee and Bryk 1986) is accurately represented. The core findings stand on the paper's data, with no external citation doing essential work. Minor issues: essentialist accounts are dismissed as lacking scholarly credibility without citing any critique, and the paper does not engage with newer British studies that might show whether the 1970s patterns persisted.

No financial or institutional conflict is identifiable. The data is publicly funded, the outcomes are administrative records, and the conclusions do not uniformly favor any political camp — single-sex schooling helped girls but not boys overall, had no lasting effect on attainment levels, and reduced gendered subject patterns without eliminating them. The structural picture is clean: no industry has a financial stake in these findings, and the results cut against the preferred narratives of multiple advocacy positions.