Part of Does single-sex schooling improve academic performance?

Does single-sex schooling improve girls' academic performance compared with coeducation, after accounting for which girls and families choose single-sex schools?

Probably not

Summary:

The small advantages seen in raw comparisons between single-sex and coeducational schools largely disappear once you account for the fact that single-sex schools tend to be private, religiously governed, or otherwise distinctive institutions. No study has cleanly shown that removing boys from girls' classrooms improves girls' test scores, and the entire evidence base comes from a handful of specific settings — 1970s Britain, Seoul, Trinidad and Tobago — that may tell us little about single-sex schooling elsewhere.

Answer in brief

Girls at single-sex schools score slightly higher on exams in raw comparisons, but this advantage shrinks or vanishes in every study that accounts for school type — private versus public, religiously governed versus secular, academically selective versus open. The one study that isolated the classroom-level effect of removing boys found a signal in girls' language scores, but it rested on just four schools. The current evidence points toward no meaningful benefit from gender separation itself, though it comes from a narrow set of countries and eras.

What the evidence shows

The raw advantage is real but misleading. In Seoul, where a government lottery randomly assigned students to schools, girls at single-sex schools scored slightly higher on the national college entrance exam — equivalent to moving from the 50th to roughly the 52nd percentile (Dustmann 2018). When Dustmann controlled for teacher gender ratios and private-school status, the advantage disappeared. Roughly 80 percent of Seoul's single-sex schools are private, compared with 30 percent of coeducational schools (Park 2012). An earlier study of the same lottery (Park 2012) reported larger advantages but could not separate gender segregation from private governance — its effects shrank 40–57 percent with even crude controls.

The meta-analytic picture agrees. Pahlke 2014 pooled 184 studies covering 1.6 million students across 21 countries. When studies controlled for which students attend single-sex schools, the average differences in math, science, and reading were close to zero for both sexes. Studies without such controls consistently showed larger advantages, confirming the apparent benefit comes from selection.

In 1970s Britain, the pattern was similar. Sullivan 2010 followed a national birth cohort and found that girls at single-sex schools had roughly twice the odds of reaching the top exam benchmark at age 16, after controlling for family background and prior test scores. But this advantage did not persist: no detectable difference appeared in exam results at age 18, university degrees by age 33, or adult skills. Sullivan did find one durable pattern — girls at single-sex schools were more likely to succeed in science and math, extending into the fields they qualified in by age 33. Whether this reflects the school environment or unmeasured family characteristics (like parents with less traditional gender attitudes choosing single-sex schools) cannot be determined from the data.

The closest anyone came to isolating the classroom effect produced thin evidence. Dustmann 2018 exploited Seoul schools converting from single-sex to coed one grade at a time. Girls who attended a newly coed school but still had all-female classmates showed no score drop, while girls who actually had boys in their classes scored lower in language. This is the right comparison, but it rests on just four converting girls' schools — too few for standard statistical methods to produce reliable conclusions. The direction is plausible, but the precision should not be trusted.

Trinidad and Tobago tells a compatible story. Jackson 2012 used a government algorithm that partially determined school assignments and found that most of the apparent advantage of single-sex schools was driven by higher-achieving students selecting into them. After accounting for selection, the benefits shrank to near zero for most girls. About 15 percent of girls who strongly preferred single-sex schools showed large gains, but this came from a tiny slice of the sample, with no way to separate gender segregation from the specific religious all-girls schools these students attended.

What is claimed but not shown

Lee 1986 claimed that girls especially benefited from single-sex Catholic high schools, but 87 percent of the girls' schools were run by religious orders versus 8 percent of coeducational schools — a confound the paper called inseparable. The study also lacked any measure of girls' academic ability before high school.

Park 2012 claimed Seoul's lottery proved single-sex schools cause better outcomes for both sexes, but its balance check used students from across South Korea rather than Seoul, and it never published an analysis of teacher gender composition that it mentioned "substantially reduced" the effect for boys' schools.

Halpern 2011 claimed single-sex schooling increases gender stereotyping, but no cited study measured stereotyping in students who actually attended single-sex schools. The evidence came from laboratory experiments on children sorted by t-shirt color.

Sullivan 2010 claimed that coeducation caused students to conform more to gender stereotypes, but the observational design cannot establish this — the association between school type and subject choice is real, the causal direction is not.

What would settle it

Randomly assigning girls to single-sex and coeducational classrooms within the same school, removing institutional differences entirely.
Reanalysis of the Seoul conversion data using statistical methods designed for very few treated units, to determine whether the classroom-level effect holds up.
Evidence from settings that resemble the countries where this debate is live — the U.S., U.K., and Australia today — rather than 1970s Britain, a Korean city with a unique lottery, or a Caribbean nation with a centralized assignment algorithm.
Studies that measure what happens in the classroom (teacher behavior, peer dynamics, stereotype threat) rather than just test scores, to identify whether a mechanism exists.

Limitations of this evidence base

The evidence comes from three settings — Seoul, Trinidad and Tobago, and 1970s Britain — each with an educational system unlike the ones where single-sex schooling is debated today. Seoul assigns students by lottery to schools with a uniform national curriculum. Trinidad and Tobago uses a centralized algorithm, and 91 percent of its single-sex schools are run by religious bodies. Britain in the 1970s had strict catchment-area rules and no published school rankings. None of these resemble the U.S. or contemporary U.K., where single-sex schooling is a choice made by specific families in a market of school options.

The time periods are also limiting. Sullivan's cohort entered school in 1969. The Seoul data run through 2009. No study in this set uses data from the past decade.

Missing entirely: large-scale evidence from U.S. public schools, studies of single-sex classrooms within coeducational schools (as opposed to entire single-sex schools), and evidence on ethnic minority girls, whom Pahlke 2014 flagged as unstudied. The strongest available designs all come from systems where single-sex schools are bundled with private governance or religious management — exactly the confound that makes the question hard to answer.

The papers behind it

Every paper this answer rests on, with the verdict from its Litmus review. The answer gives weight only to claims those reviews found supported.

Where the claim comes from

Effects of single-sex secondary schools on student achievement and attitudes.Lee and Bryk (1986). Journal of Educational PsychologyDubious — probably doesn't hold up
Claims that single-sex Catholic high schools produce better academic outcomes than coeducational ones, especially for girls. The strongest findings are for girls' schools, but 87% of those schools were run by religious orders versus 8% of coeducational schools — a confound the paper itself calls inseparable — so the results cannot distinguish gender separation from a different style of institutional management.

Original studies

Single-Sex Schooling and Academic Attainment at School and Through the LifecourseSullivan et al. (2010). American Educational Research JournalCredible — likely right, with caveats
Using a British birth cohort from 1958, this paper finds that girls at single-sex schools scored higher on national exams at age 16, and that both sexes pursued more gender-atypical subjects (girls in science, boys in languages) at single-sex schools — a pattern that persisted into their thirties. The associations are well-measured and the controls are strong for an observational study, but the paper's causal claims — that co-education intensified gendered patterns and that single-sex environments reduced stereotyping — go beyond what the design can establish.
Single-sex schools, student achievement, and course selection: Evidence from rule-based student assignments in Trinidad and TobagoJackson (2012). Journal of Public EconomicsCredible — likely right, with caveats
Uses Trinidad and Tobago's centralized school-assignment algorithm to estimate the causal effect of single-sex schooling — a genuinely clever design that convincingly shows naive comparisons are inflated by selection. The core finding that most students show no benefit is credible, but the more specific claims — that girls with strong preferences benefit enormously, that single-sex schools reduce girls' science-taking, and that universal expansion would be pointless — rest on small subgroups, untested assumptions about course offerings, and extrapolations beyond what the data support.
Causal Effects of Single-Sex Schools on College Entrance Exams and College Attendance: Random Assignment in Seoul High SchoolsPark et al. (2012). DemographyDubious — probably doesn't hold up
Claims that Seoul's lottery-based school assignment proves single-sex schools cause better exam scores and college attendance for both boys and girls. However, 80% of single-sex schools are private while only 30% of coeducational schools are, and the estimated effects shrink 40–57% with even limited controls — meaning the paper cannot separate the effect of gender-segregated classrooms from the effect of private-school governance, teacher hiring, and institutional culture.
Why Are Single-Sex Schools Successful?Dustmann et al. (2018). Labour EconomicsCredible — likely right, with caveats
Students randomly assigned to single-sex schools in Seoul scored higher on college entrance exams than those in coed schools, but this advantage disappeared once other school characteristics were controlled for. The paper's most distinctive claim — that mixing boys into girls' classrooms specifically harms girls' language scores — rests on just four converting schools analyzed with statistical methods known to be unreliable at that scale, so the direction of the finding is plausible but the claimed precision should not be taken at face value.

Studies that pool other studies

The effects of single-sex compared with coeducational schooling on students’ performance and attitudes: A meta-analysis.Pahlke et al. (2014). Psychological BulletinCredible — likely right, with caveats
This meta-analysis of 184 studies finds that when studies control for selection effects, the average difference between single-sex and coeducational schooling is small across academic and social outcomes. The directional finding is probably right, but the strong conclusion that single-sex schooling provides 'no benefits' overshoots the evidence — the 'controlled' category mixes strong and weak designs, the 12 truly randomized studies are never analyzed separately, and pooling fundamentally different types of single-sex programs could mask real effects that cancel each other out.

Critiques

The Pseudoscience of Single-Sex SchoolingHalpern et al. (2011). ScienceDubious — probably doesn't hold up
This policy essay argues that single-sex schooling has no proven academic benefits and harms children by increasing gender stereotyping, calling for federal regulations to be rolled back. The argument demands rigorous experimental evidence before accepting that single-sex schools help, but accepts laboratory experiments and preschool correlational data as evidence that they harm — an asymmetry that aligns with all seven authors' roles as founders of an advocacy organization for coeducation, and that undermines the paper's claim to be following the evidence wherever it leads.

Cited but not reviewed

The answer mentions these, but Litmus hasn't reviewed them, so they carry no weight here: In 1970.