BLOG: The Impact of Single-Sex Education on Educational and Personal Outcomes

By Caitlin McCarey. Wellesley College

This review assesses the impact of single-sex education programs on educational and personal outcomes of lower, middle, and high school aged students. This discussion is guided by two research questions, grounded within an economic framework. First, are single-sex education programs more beneficial for students than co-educational (co-ed) programs? Second, are the effects of single-sex education programs heterogeneous for different groups (boys vs. girls, majority vs. minority students, and high vs. low income students)? Quality randomized controlled trials have only recently been utilized to answer these questions. The review concludes that single-sex schools range from “equally as effective” to “more effective” than co-ed schools along measures of students’ personal and academic outcomes. Therefore, we recommend that a program of increased school choice where families can choose between single-sex and co-ed schools.

Advocates of single-sex education trumpet the many benefits of learning in a single-sex environment: girls become more confident in their abilities, boys embrace their emotions, classroom disruptions are reduced, and educational outcomes are increased for everyone (Hayes et. al., 2001; Hughes, 2007). Opponents of single-sex education, however, claim that segregating the sexes vindicates pseudoscience and will be outright harmful because it encourages hyper-sexuality and reinforces the gender binary (Eliot, 2007; Halpern et. al., 2011; Goodkind, 2012). Until recently, the dearth of high quality randomized controlled trials allowed the propagation of misinformation regarding single-sex education.

This paper will examine two meta-analyses that conducted a review of the single-sex education literature within the past decade. Very recent studies, including randomized controlled trials in Korea and random experiments in the UK, will then be described and analyzed to address holes in these meta-analyses. Overall, this paper will conclude that while approximately two-thirds of studies find no difference between single-sex and co-ed schools, approximately one-third of studies find advantages to attending a single-sex school. To my knowledge, no studies have found that single-sex schools harm students or that co-ed schools benefit students. The literature includes studies on all-boys and all-girls environments in lower, middle, and high schools.

This literature review will proceed in several steps. First, I will describe the importance of researching single-sex schools. Second, I will describe the economic theory behind the general equilibrium that arises after the introduction of single-sex schools to a school district. This discussion yields the following two research questions: (1) Are single-sex education programs more beneficial for students than co-educational programs? (2) Are the effects of single-sex programs heterogeneous for different groups (boys vs. girls, majority vs. minority students, and high vs. low-income students)? Third, I will provide an overview of the empirical evidence of these policy questions. Fourth, I will examine the causal mechanisms that may have lead to these results. Fifth, I will assess the external validity of the studies discussed in the third section. Sixth, I will briefly discuss the cost-effectiveness of single-sex education programs. Finally, I will provide recommendations for future experiments and conclude by answering the research questions.

Section One: The Importance of Single-Sex Education

It has been shown that girls and boys enter kindergarten with the same aptitudes and same abilities in all subjects—including science and math. However, over the six years of elementary school, girls lose two-tenths of one standard deviation in test scores relative to their male peers (Fryer & Levitt, 2009). This is half of the black-white test score gap over the same period (Fryer & Levitt, 2009). If social conditioning or societal attitudes about gender roles are holding girls back in the classroom, single-sex schools may present a solution. After all, an individual’s education level impacts her likelihood of attending college, and choice of major and success in college will influence her choice of career and future earnings. There is great inequity in the workplace. The gender wage gap persists across all industries, and men are still much more likely than women to earn a degree in business and STEM.

Therefore, increasing women’s educational achievement is a key goal for the United States and the world. Higher levels of education have been linked with a plethora of desirable characteristics, such as lower infant mortality rates, lower teen birth rates, higher levels of economic output, higher wages, and greater gender equality. Especially in countries were co-ed schooling is culturally or religiously undesirable, single-sex schools may provide the solution to increasing overall education levels for girls and women. If single-sex schools increase girls’ competitiveness and confidence levels in their abilities, single-sex schools may contribute to the reduction of both the achievement-gap and the gender wage gap in Westernized countries like the United States.

After the No Child Left Behind (NCLB) Act was modified by Congress in 2006, single-sex schools were no longer considered a violation of Title IX. This led to a 4,400% increase in the number of single-sex classrooms in the United States by 2010 (Booth et. al., 2013). It is important for U.S. policymakers to study single-sex environments (both single-sex classes and single-sex schools) as more and more students are opting to learn in them. Currently in the US, single-sex schools are seen as the purview of the wealthy or the religious, but this may change as more and more studies on the benefits of single-sex schools are released. In order to stay on top of these trends and to ensure the best educational programs for its students, the U.S. must study these programs rigorously.

Section Two: Economic Theory and Intuition

It is not clear how the addition of single-sex classrooms or single-sex schools would affect education production in a general equilibrium. This is because single-sex education is not universally desired. There are many families who, due to misinformation or personal reasons, may not want their children to attend a single-sex school. For example, while many families view single-sex schools as empowering and cutting-edge, an equal or larger number may view them as outdated anachronisms of a more sexist American educational system. Additionally, if single-sex schools are private or religious schools—as they overwhelmingly are in the U.S.—families could have additional reasons to select or not select into a single-sex school. These reasons could include both location and income constraints. Therefore, while the addition of a single-sex school to a school district may be most analogous to the addition of a private school in terms of the general equilibrium effects, we cannot be certain this is an accurate comparison.

Let us proceed with the assumption that families prefer single-sex schools is over co-ed schools. This means, if a single-sex school enters a school district, most families would prefer to send their children to the single-sex school instead of the public co-ed schools. Under this assumption, the entrance of a single-sex school to a school district will function as an exogenous shock on the education system. The equilibrium result could be increased competition as the public schools attempt to compete with the single-sex school. This could lead to an increase in the resources and average test scores in the public schools, improving the delivery of education services. However, there could also be public school decay as the smartest, wealthiest students leave the public school system to attend the single-sex school.

Moreover, the addition of a new school would drive up the demand for teachers, leading either to a salary increase or the departure of the best teachers from the lower-paying sector to the higher-paying sector (the higher-paying school will often be the co-ed public school). The effect could be reversed, however, if teachers prefer teaching in single-sex environments or if the single-sex school is associated with smaller class sizes, brighter students, wealthier parents, a certain racial composition, or other characteristics upon which teachers make their educational decisions.

If the single-sex school will only accept boys or only accept girls, this could lead to a gender imbalance in the public schools. If having more girls in a classroom is beneficial for all students and having more boys is detrimental, there could be additional effects on the students in the public school. Finally, if families prefer single-sex schools to co-ed schools, families who derive higher utility from sending their children to such schools could move into the school district. This could drive home prices higher, which could benefit the public school system if they are funded mainly from property taxes. It is unlikely families would flee the district if they did not have a preference for single-sex schooling. Therefore, there could be important general equilibrium concerns for school districts contemplating the addition of a single-sex school.

Section Three: Causal Evidence

It is difficult to quantify the effects of single-sex classrooms and single-sex schools for several reasons. First, students self select into these environments. That is, students with nonrandom characteristics are more likely to choose to attend single-sex schools. Family characteristics like income or religion could affect this choice along the extensive margin (insofar as many single-sex schools are Catholic schools or private schools), while the student’s personal characteristics and abilities could affect the decision along the intensive margin. Additionally, single-sex schools are overwhelmingly private schools.

The effects of learning in a private school (which traditionally has smaller class sizes and more resources than public schools) may obscure the true effect of learning in a single-sex environment. It is difficult to tease out these two effects. Therefore, the studies with the highest internal validity will involve randomized controlled experiments that capitalize on situations where students are randomly assigned to single-sex vs. co-ed schools or classrooms, or where curricula, teachers, and resources are fixed across classrooms and schools. Several studies meet these criteria; they will be discussed late in this section. But first, the results of two meta-analyses will be examined to provide an overview of the single-sex schooling landscape.

The first meta-analysis I will review was conducted by economists Erin Pahlke, Janet Shibley and Carlie Allison (2014). They analyzed data from over 180 studies conducted in 21 nations on the effects of single-sex schooling on academic outcomes and student characteristics (such as confidence, competitiveness, etc). In order for a study to be included in their meta-analysis, it had to 1) contain quantitative data on student outcomes, 2) assess K-12 schooling, 3) measure personal or academic outcomes, 4) include separate treatment and control groups (Pahlke, Shibley, & Allison, 2014). Out of almost 2,500 studies regarding single-sex education, only 184 studies found to meet these criteria.

These 184 studies were then labeled as “controlled” or “uncontrolled” studies. A controlled study used random assignment of students to single-sex or co-ed classes/schools or controlled for family characteristics like socio-economic status. Uncontrolled studies had no controls for selection effects. Therefore, the controlled studies possess the strongest research methods. Table 1 is a visual representation of the results obtained by the 57 controlled studies; it is modified from a table of all 184 studies in Pahlke, Shibley, & Allison (2014).  

McCarey_table_2017

Pahlke et al. (2014) separated the results of their meta-analysis by effect category (i.e. academic outcomes, personal outcomes, and so on), so that is how I will proceed in my review. Due to their weaker research design, uncontrolled studies will not be examined here. The results presented here are average weighted effects, which gives greater weight to studies with larger sample sizes. First, the authors examine the effects of single-sex education on students’ math scores. The overall effect of single-sex education on students’ math scores was positive but very small (Pahlke, Shibley, & Allison, 2014). There were larger effects for male students than for female students, but because the positive estimates are both so small, the authors conclude there is a negligible effect of single-sex schooling on student math scores. There was some inter-group heterogeneity: girls in middle school derived modest improvements in math scores from placement in single-sex environments (Pahlke, Shibley, & Allison, 2014).

Next, the authors examine the effect of single-sex classes and schools on students’ mathematics attitudes, concluding that there is no effect for either male or female students (Pahlke, Shibley, & Allison, 2014). There was no weighted effect of single-sex schooling on students’ performance in science, and there were not sufficient studies to conduct an analysis on the change in students’ scientific attitudes. The effects of single-sex schooling on verbal performance were also close to zero (Pahlke, Shibley, & Allison, 2014). In terms of general school achievement, the difference in female achievement was small but positive; there were not enough studies to measure the effects on male students (Pahlke, Shibley, & Allison, 2014).

Additionally, there were not enough studies to determine the effects on students’ attitudes towards general education. The weighted effect of single-sex on gender attitudes, however, implied that female students in co-ed schools were much more likely to endorse gender stereotypes than single-sex cohorts. However, the authors urge caution in interpreting this effect due to variance between the weighted and unweighted effects. There were insufficient studies to measure the effects of single-sex on male stereotypes. There was no difference in student aspirations across single-sex and co-ed schools.

Overall, the authors conclude that there are only trivial advantages to attending a single-sex school versus a co-ed school along important measures like student achievement and student attitudes (Pahlke, Shibley, & Allison, 2014). The authors attribute this dissonance between the perceptions of single-sex schooling benefits and these studies to the dearth of rigorous study design. They posit that many studies do not control for selection effects and student characteristics, which biases public opinion regarding the results. Additionally, the authors conclude that there are not heterogeneous affects by gender, socio-economic status, or age. These results contradict studies that have found large positive effects of single-sex schooling for younger girls. Therefore, they conclude that there is no evidence that single-sex schools offer educational and personal advantages to either male or female students.

A second meta-analysis was conducted by Fred Mael, et al. (2005) at the behest of the Department of Education. This study uses the evaluation standards of the Campbell Collaboration (CC) and the What Works Clearinghouse (WWC) to evaluate studies on the effects of single-sex education. Overall, they state their results are equivocal—there is no evidence of either benefit or harm from attending a single-sex school. To be included in this meta-analysis, the studies had to involve students enrolled full-time in elementary, middle, or high schools in Westernized or English-speaking countries. As with Pahlke’s meta-analysis, no studies involving college students were included. The condition on Westernized countries was added to ensure the results were applicable to the U.S. education sector.

When the strict guidelines of the WWC were applied to the single-sex education literature—that is, that all studies other than randomized controlled trials, quasi-experimental designs with matching, or regression discontinuity designs be excluded—no studies remained for analysis. Therefore, Mael et al. modified the WWC standards and included all correlational studies utilizing robust statistical controls for student ability, family SES, and school resources (Mael et al., 2005). There were 40 quantitative studies that met the modified criteria. As Mael’s meta-analysis was published before Pahlke’s, all of the 40 studies examined by Mael are included in Table 1, which is based off of a summary table appearing in Pahlke et. al. (2014). Mael’s results are either pro-single-sex schools or show no difference between outcomes among graduates of single-sex and co-ed schools.

 

The results of Mael et al.’s study are organized along several classifications. First, they state that most studies report either positive effects or no effects of single-sex schools on students’ academic achievement. Specifically, of studies that examined students’ performance on mathematics, science, English, and social sciences tests, roughly a third found that students in single-sex environments outperformed students in co-ed environments (Mael et al., 2005). The remaining two-thirds of studies found no difference in academic performance. These results argue moderately in favor of single sex schools. Next, the authors state that long-term measures of student success (such as college attendance or graduation rates) cannot be directly attributable to single-sex schools. Recent research conducted in Korea has found positive effects of single-sex schools on college attendance (Park, Behrman, & Choi, 2012b). These studies will be discussed later.

The next category of results that Mael et al. analyze involves student socioemotional development. Overall, the results are mixed. Interestingly, a third of studies regarding self-esteem found positive effects for co-ed schools, while the remainder showed no preference. These positive effects were only for male students. Additionally, the authors state that a majority of studies found that single-sex schools had a positive effect on the higher educational aspirations of female students. There were positive effects on career aspirations for boys at single-sex schools, but not for educational aspirations (Park, Behrman, & Choi, 2012b). The authors write that there were not enough studies to issue blanket statements regarding the effects of single-sex schools on overall gender inequity, school climate, unemployment, graduation rates, mental health, or parent and student satisfaction. Overall, Mael et al. agree with Pahlke et al. that there is a dearth of quality, randomized studies with robust statistical controls (Mael et al., 2005).

Since the publication of these two meta-analyses, several studies with more rigorous experimental designs have been written. These have studied the effects of random assignment of students in metropolitan areas to co-ed or single-sex middle and high schools in South Korea (Park, Behrman, & Choi, 2012a; Park, Behrman, & Choi, 2012b; Lee, Niederle, & Kang 2014a; Lee, et. al, 2014b) or the random assignment of students to co-ed vs. single-sex classrooms within co-ed universities (Booth, 2013; Booth and Nolen, 2012; Eisenkopf, 2014). While many of these studies found positive effects of single-sex schooling, a non-inconsequential number found no difference between single-sex and co-ed schools. However, based on our prior analysis of the Pahlke and Mael meta-analyses, this is not surprising. These studies will be examined below.

The Korean studies take advantage of a random assignment program that was implemented in 1974 and occurs in Seoul and the six largest metropolitan centers in Korea. Students are randomly assigned via a lottery system to either co-ed or single-sex middle and high schools within their district; almost 75% of Korean high school students attend schools that are included in the lottery system (the remaining 25% attend vocational high schools). Importantly, both public and private schools partake in this lottery system and cannot reject the students assigned to them.

While non-compliance in such a program would traditionally be a key concern, Park et a. argue the potential bias from non-compliance and attrition are minimal. This is because the only way for students to transfer to another school is to move to a new school district. In this new district, students would be reassigned to a new school based on the same random lottery, thereby minimizing or negating the benefits of moving in the first place. Additionally, Park et al. showed that only a very small number of households do move during the relevant timing of high school, indicating that non-compliance is not biasing the estimates (Park, Behrman, & Choi, 2012a).

The Korean studies measure the effect of single-sex schools on student test scores, college entrance exam scores, rates of college attendance, choice of college major and career, and student competitiveness. Park, Behrman, and Choi conclude that attending a single-sex high school is associated with higher average scores on Korean and English tests, higher rates of four-year college attendance, and lower rates of two-year college attendance for both male and female students (Park, Behrman, & Choi, 2012b). The increases in test scores are large (7% to 15% of one standard deviation), although not all are statistically significant.

Park, Behrman, & Choi (2012b) also found that male and female students at single-sex schools are more likely to attend four year colleges, and less likely to attend two year colleges, than co-ed cohorts. The differences in college attendance are quite substantial—.5 standard deviations for girls and .8 standard deviations for boys (Park, Behrman, & Choi, 2012b). These results are robust to controls for school-level variables such as teacher quality, the student-teacher ratio, and the proportion of students receiving lunch support, and whether the single-sex schools are public or private (Park, Behrman, & Choi, 2012b).

In a second study of the Korean high schools, Park, Behrman, and Choi obtain additional administrative data and longitudinal survey data on students’ college entrance exam math scores and choice of a STEM college major two years after enrolling in college. Park et al. found that both male and female students attending single-sex schools outperformed their co-ed counterparts on national standardized math tests. The positive effects of attending a single-sex high school were quite large (ranging from 7% to 25% of one standard deviation, and were larger for boys than for girls. Interestingly, despite the higher math test scores, female students at single-sex high schools were no more likely to major in STEM in college than female students from co-ed high schools. However, boys from single-sex high schools were both more likely to choose a STEM major and more likely to report a high interest in STEM subjects. These patterns held in both Seoul and non-Seoul schools.

There are several internal validity concerns that arise in studies that measure the effects of single-sex schools in Korea. First, after concerns arose over lack of school choice, some districts outside the capital modified the random lottery to allow students to list two or three schools are their “preferred” choice. In each non-Seoul school, 30-40% of their enrolled students were randomly selected from the students who had labeled that school one of their preferred schools, while the remaining 60-70% of students were determined completely at random (Park, Behrman, & Choi, 2012a). Due to this policy modification, some of the students in each school are able to self-select into that school. This introduces selection bias that may prejudice the results. However, after comparing the characteristics of the Seoul and non-Seoul cohorts, Park et al. found that students at single-sex schools and co-ed schools had similar levels of academic achievement socio-economic backgrounds.

A second internal validity concern for the Korean studies is that public and private school teachers are subject to different licensing requirements and recruitment incentives. For example, public school teachers have to take additional teaching exams and are subject to school rotations every 4-5 years. As Park et al., states, this may lead to differences in teacher-related characteristics across public and private schools. Finally, the majority of single-sex schools in Korea are private, while most of the co-ed schools are public. This means that in addition to varying along the lines of gender composition, single-sex schools and co-ed schools could also vary along other important characteristics such as teacher quality or characteristics.

Other Korean studies have taken advantage of the random assignment of middle school students to co-ed middle schools, co-ed middle schools with single-sex classrooms, or single-sex schools (Lee et al, 2014; Lee, Niederle, & Kang, 2014). Middle school students in Korea are assigned to their schools based on a similarly random procedure as governs high school assignment, except no middle school students can submit school preferences. Compliance is high, because the only way to change to another school is to move to another school district and re-enter that district’s school lottery. These studies aim to counteract some of the biases that may be influencing the high school studies. These studies have high internal validity because students are assigned to classrooms in a process that equalizes prior academic achievement, and there is strict adherence to a nationally designed curriculum. Additionally, all schools receive equal funding, which removes concerns over differences in resources.

Lee et al. (2014) conclude that male students attending single-sex schools outperform male students in co-ed classes by .15 standard deviations. They state that this improvement was driven by increases in male students’ effort and time spent studying. Interestingly, male students in single-sex classes within co-ed schools perform .10 standard deviations below male students in co-ed classes. Lee et al. concludes that this achievement gap between boys in co-ed schools but single-sex classrooms and boys in co-ed schools and co-ed classrooms is due to the change in gender composition. There were no differences in test scores across the female students in co-ed vs. single-sex classes. They offer no explanation for this result.

A final study uses the random assignment of Korean students to middle schools to measure the effect of the gender composition of classrooms on students’ willingness to participate in competitive activities. Lee, Niederle, and Kang (2014) conclude that, contrary to popular belief, the gender gap in competitiveness is not reduced by single-sex schooling. This result is robust to controls for student achievement and parent characteristics. To measure competitiveness, the authors created various competitions where students competed against each other to solve arithmetic problems. While there are no differences in student performance on the tasks, female students from single-sex schools are 15 percentage points less likely to opt into a tournament-style (i.e. highly competitive) game than are boys in co-ed schools (Lee, Niederle, & Kang, 2014). Girls in co-ed middle schools are only 8 percentage points less likely to choose to compete in the tournament (Lee, Niederle, & Kang, 2014). Several randomized experiments conducted at a University in the United Kingdom disagree with this conclusion.

Additional studies have conducted randomized experiments on students at co-ed universities. Booth et al (2013) designed an experiment where all first year economics students at the University of Essex in the United Kingdom were sorted into either single-sex or co-ed discussion sessions for their ECON 101 classes. The classes were year-long and taught for two hours every week by one professor in a lecture hall, supplemented by one hour of instruction per week taught by teaching assistants (TA). Students were randomly assigned to co-ed, all-boy, or all-girl TA classes. The co-ed and single-sex classes were the same on almost all observable characteristics, and there was no attrition because students were not allowed to switch classes.

Booth et al. found that female students in the single-sex classes were 7.5% more likely to pass their first year courses and scored 10% higher in their required second year economics courses than their female peers assigned to the co-ed class. Additionally, women in single-sex classes were 9.3 percentage points more likely to complete optional homework assignments and nearly 6 percentage points more likely to attend classes. However, women in the single-sex classes were no more likely to take technical classes (econometrics or advanced math courses) in later years than their counterparts. There was no affect on the passage rates or test scores of male economics students.

A second study conducted concomitantly by Booth et al. (2013) with the students at Essex University included an experiment designed to test students’ risk tolerance. They administered tests to all economics students in the first week and the eighth week of class. While female students were more risk averse than male students at both times, female students in the single-sex TA-classes were significantly more likely to exhibit risk-seeking behavior in the 8th week than female students in the co-ed TA-classes. The magnitude of this effect was substantial. There was no change in risk aversion for male students. These results were robust to the inclusion of controls for IQ, academic ability, and personal characteristics. A second but comparable study by Booth and Nolen (2012) found that middle school girls who attended public single-sex schools exhibited similar levels risk tolerance to boys from co-ed schools in controlled experiments. Both of these studies disagree with the results of Lee, Niederle, & Kang (2014) that attending single-sex schools does not affect the competitiveness of female students.

Section Four: Evidence on Causal Mechanisms

In order to design the most effective education policy, it is imperative that we understand why single-sex schools might produce different outcomes than co-ed schools. A plethora of possible mechanisms have been discussed in the literature. These mechanisms include: 1) tracking; 2) peer effects; 3) decreased distractions during class; 4) increased number of same gender role models; 5) differences in curriculum/classroom organization; 6) differences in funding/resources; 7) reduction in stereotype threat. Each of these mechanisms will be discussed, although the relative weight and importance of each mechanism cannot be determined due to the difficulty of separating their effects.

The first explanation for why single-sex schools could effect improvement along academic dimensions is the model of student tracking. This model posits the idea that students will perform better if they are surrounded by peers with similar academic abilities. For example, Duflo et. al. (2011) found that sorting students into elementary school classes based on ability caused both high-achieving and low-achieving students to perform better on end of the year tests. If students have similar abilities—say, in an honors class where all students are above average—the teacher will be able to teach to the median level of achievement, which will be a satisfactory level for the majority of students. In a class with a large distribution of academic achievement may lead teachers to teach to the right tail of the achievement distribution. If teachers are advancing through the material too quickly for the majority of students in the class, the removal of the top students may allow the teacher to slow down, and therefore teach at a more appropriate level.

If female students are less likely to be in the right tail, isolating students by gender may reduce the variance in female student achievement and allow a tracking mechanism to improve student scores. Alternately, if female students are more likely to be the top performers in a class, removing male students may allow the pace of an all-female class to accelerate. However, Booth et. al (2013) reject tracking as the mechanism through which single-sex schools benefit students. They demonstrate that both the distribution and standard deviation of ability are the same for both female-only and male-only classes. That is, both male and female students have similar ability distributions. This implies there will be no tracking by ability if one segregates classes based on gender. Therefore, it seems unlikely that tracking is the cause of the benefits of single-sex schooling.

A second mechanism through which single-sex schools could benefit students is peer effects. Peer effects include any way in which a student’s classmates affect his or her behavior, such as through peer instruction, knowledge spillovers, influence on classroom standards and disciplinary behaviors, and overall classroom environment (Hoxby, 2000). Often it is very difficult to isolate various peer effects. There are some ideas how peer effects could be exerting influence on students in single-sex environments. First, it is traditionally stated that students in single-sex environments care less about superficial qualities like looks or popularity and spend more time focusing on schoolwork.

Lee et. al. found that female Korean middle school students in single-sex schools were more likely to report that their peers worked hard than female students in single-sex classes within co-ed schools, and that male students at single-sex schools spent more time studying than male students at co-ed schools (Lee et. al., 2014, Park et. al., 2012b). There could also be peer effects from increasing the number of female students in a class. Several studies have found that both male and female students perform better as the fraction of female students rises (Lavy and Schlosser, 2011; Hoxby, 2000) and that female students complete more optional assignments when assigned to single-sex classes (Booth et. al., 2013). Therefore, it seems that peer effects could be one potential mechanism through which single-sex increase student outcomes.

A third mechanism explaining these benefits could be that female students in single-sex environments may be subject to fewer distractions, disturbances, interruptions, or violent episodes than female students in co-ed classes. This could be because young boys are socialized to respond differently in academic environments, or because of genetic differences in boys’ brain chemistry. Whatever the cause, several studies have found that decreasing the number of boys in a classroom increases both boys’ and girls’ achievement. For example, when Lavy and Schlosser (2011) concluded that increasing the fraction of female students in a classroom raised overall academic performance of all students, they concluded that the mechanism through which this effect operated was a reduction in classroom disruption and violence, and not due to spillover effects (Lavy and Schlosser, 2011). They found no change in individual students’ behavior, implying that the positive effects operated through a compositional change, i.e. gender peer effects. Hughes (2007) also found that the gains from single-sex schooling come in part from decreased classroom disturbances and greater equity of opportunity for female students.

A fourth mechanism could be the increased presence of same-gender role models at single-sex schools. It has been shown that single-sex schools employ a greater number of same-sex teachers—all-boys schools employ more male teachers than co-ed schools, while all-girls schools employ more female teachers (Riordan, 1990). Studies have shown that female high school students achieve higher levels of schooling if their co-ed high schools have a higher fraction of female professional staff. If students learn more effectively and efficiently from teachers of their own gender, as has been posited, this effect could be one of the reasons for increased performance at single-sex schools.

While the economics of education literature has shown that minority students perform better with a minority teacher (Dee, 2004), the results are less conclusive regarding gender. For example, Park et. al. (2012b) proposed that one of the explanations for the vast gains in English and Korean test scores experienced by both boys and girls in single-sex schools could be attributed to the higher numbers of same-sex teachers. They cite research that shows higher fraction of female teachers could counteract research that shows that boys receive higher levels of encouragement and reinforcement, especially in math and sciences, than female students (Park et. al., 2012b). However, Lee et. al. (2012) concluded that while single-sex schools were better for both male and female students, this effect was not driven by the gender composition of the schools. More research must be conducted on this subject.

I will combine the discussion of the next two mechanisms, differences in classroom organization and funding levels, because they are so interrelated. In the US, most single-sex schools are private schools, and most co-ed schools are public schools. Therefore, it is very important to ensure the results of studies that purport to show the effects of single-sex schools are not mistakenly reporting the effects of being in a private school. Studies that were conducted in Korea, where students are randomly assigned to either single-sex or co-ed middle and high schools, provides guidance in this area. Importantly, all schools in Korea are designed around a strict national curriculum where there are very few opportunities for teachers to modify lessons or teaching methods. Second, schools in Korea are funded through a centralized process, so there are not the vast funding disparities that exist in the U.S. Because these studies found substantial benefits of attending single-sex schools holding these characteristics constant, it appears that the benefits of single-sex schools do not operate through differences in classroom organization or funding levels (Park, Behrman, & Choi, 2012a; Park, Behrman, & Choi, 2012b; Lee, Niederle, & Kang 2014a; Lee, et. al, 2014b).

Finally, the abolition of “stereotype threat” could explain the benefits of single-sex classrooms and schools, especially for female students. Stereotype threat occurs when a student intentionally restricts or modifies their participation in classroom activities to avoid making mistakes that would cause the student to fulfill negative stereotypes about their race, gender, age, etc. For example, female students in math or science classes may ask fewer questions in order to avoid giving the impression that they do not understand the material, because it is a stereotype that girls perform more poorly in math and science classes. Steele, Spencer, & Aronson (2002) found that girls reported more self-confidence in their math abilities after learning math in a single-sex classroom. However, it is extremely difficult to devise an empirical experiment or test that could ascertain whether the abolition of stereotype threat improves female students’ performance. It is also unclear whether single-sex schools for male students would be subject similar mechanisms.

Section Five: External Validity

As I have discussed throughout this paper, self-selection is the greatest threat to both the internal and external validity of studies involving single sex schools. While this bias can be mitigated or even removed through robust experiment design, these sorts of experiments have been few and far between to date. In general, if a study does not control for various school-level and student-level characteristics, we can state with confidence their results are heavily biased by self-selection and student family characteristics. These studies will not be useful as we craft policies regarding single-sex vs. co-ed education.

However, there are several other factors that may affect the external validity of the studies within the single-sex education literature. First, there are the inherent differences across the education systems in non-Western vs. Western countries that make comparisons difficult. In Korea, Thailand, Trinidad and Tobago, and Ireland, single-sex schools are much more common than they are in the US. Single-sex schools in these districts are not subject to increased participation due to their novelty. There are also many differences between these school districts and school districts in the US that makes the application of the results of these studies potentially dubious. For example, differences in teacher training requirements and payment structures may affect the quality of teacher candidates, as well as the methods they use in their classrooms. Finally, none of these countries have the racial tension that exists between minority and majority students in the U.S., so we cannot know for certain how the results of foreign studies would apply to White vs. Black/Latino/Asian students.

Additionally, due to the challenge of conducting randomized experiments with single-sex environments, many of the studies that exist draw on smaller sample sizes. Unlike other fields of economics that can use census data or state-level data, the economics of education literature (and especially single-sex education literature) lack such large-scale, long-term studies. There is as yet no “Tennessee STAR Experiment” for single-sex schools. If the sample sizes of single-sex school studies are not large enough to be representative of the population from which they attempt to sample, the results will be subject to higher variability and error. In the worst-case scenario, the positive effects of single-sex schooling could actually be due to an extremely talented cohort of students.

Finally, many of the studies examining single-sex literature have examined a variety of ways in which these environments may improve student outcomes—from academic to personal outcomes. As there are differing levels of robustness within these experiments, it can be hard to compare studies that also differ in their definition of a successful program. For example, if single-sex classrooms don’t improve student performance but increase girls’ self-confidence, are they still worth it? Additional challenges arise when we attempt to rank or otherwise choose amongst these outcomes: is self-confidence more important than GPA? Are test score increases more powerful than a decrease in classroom distractions? There is no right answer to these questions, but how we choose to answer them will have vast consequences on the studies we conduct and the policies we implement.

Section Six: Cost effectiveness

One of the greatest potential benefits of implementing single-sex education in the US would be its cost: free. No new resources would need to be provided in order to transition co-ed environments to single-sex environments. All we would need to do is reshuffle students into single-sex classrooms within co-ed schools or reassign them to single-sex schools within their school districts. The benefits of single-sex schooling appear even more substantial given the lack of capital required. Other popular educational programs—such as reducing class sizes, improving student-teacher ratios, or implementing improved teacher training programs—would require a substantial upfront and prolonged investment.

Section Seven: Recommendations for Research

As was highlighted in both of the meta-analyses, there is a dearth of rigorous, randomized controlled trials with robust statistical controls that test the effects of single-sex education. In a perfect (hypothetical) world, I would want to design an experiment that utilized random assignment and controlled for selection effects. The experiment would take place on a large scale—perhaps across several states—and occur in the U.S. to maximize its applicability to U.S. education policy. I would want to compare students across various parts of the country, perhaps in New York, Chicago, Florida, Texas, and California, to ensure the results weren’t specific to just one state or one school district. All of the students would be randomly assigned to a treatment (single-sex school) or a control (co-ed school) group.

The treatment group would be split into several further groups: students assigned to single-sex schools from kindergarten-onwards, students assigned to single-sex schools from middle school onwards, and students assigned to single-sex schools from high school onwards. Not only would this allow us to measure the effects of single-sex schools for various age cohorts, but it would also allow us to determine if the duration of exposure to the treatment affected students’ personal characteristics and educational outcomes. Of course, we would make sure all of the groups were balanced across both student characteristics (such as prior academic achievement and personal characteristics) and family characteristics (SES, educational attainment of parents, number of siblings, birth order, neighborhood characteristics). The curriculum and funding levels of all of the schools would be standardized at the national level, as in the Korean studies. A study of this nature would greatly enrich the single-sex schooling literature.

Conclusion

I will conclude by returning to the two research questions asked at the beginning of this literature review: (1) Are single-sex education programs more beneficial for students than co-educational programs? (2) Are the effects of single-sex programs heterogeneous for different groups (boys vs. girls, majority vs. minority students, and high vs. low-income students)? To answer the first question, most studies find no difference between the outcomes of students in single-sex schools and co-ed schools, while some studies find results that favor single-sex schools. No studies have found any negative effects of single-sex education on students.

Due to the paucity of literature addressing the heterogeneous effects of single-sex schools, it is difficult to answer the second question. We cannot conclude that single-sex schools are more or less beneficial for minority vs. majority students or high vs. low-income students because very few studies have attempted to separate out the effects for these groups. However, we can state that while both boys and girls appear to benefit from single-sex schooling, girls may benefit more along academic and personal measures than boys.

These results could be due to peer effects, decreases in classroom distractions, increases in female role models, or the abolition of stereotype threat. It is unlikely that the benefits of single-sex schools are due to tracking, differences in classroom organization, or school funding levels. Overall, I recommend a policy of increased school choice where parents may decide whether to send their children to a single-sex school or a co-ed school.

References

Booth, Allison, and Patrick Nolen. 2012. “Choosing to Compete: How Different are Girls and Boys?” Journal of Economic Behavior & Organization 81: 542-555.

Booth, Allison, Linda Cardona-Sosa, and Patrick Nolen. 2013. “Do Single-Sex Classes Affect Achievement? A Study in a Coeducational University.” Borradores de Economía Núm. 787: 1-23.

Booth, Allison, Linda Cardona-Sosa, and Patrick Nolen. 2014. “Gender Differences in Risk Aversion: Do single-sex environments affect their development?” Journal of Economic Behavior and Organization 99: 126-154.

Brown, Charles, and Mary Corcoran. 1996. “Sex-Based Differences in School Content and the Male/Female Wage Gap.” Working Paper 5580. Cambridge: National Bureau of Economic Research.

Dee, Thomas. 2005. “Teachers and the Gender Gaps in Student Achievement.” Working Paper 11660. Cambridge: National Bureau of Economic Research.

Dee, Thomas. 2004. “Teachers, Race, and Student Achievement in a Randomized Experiment.” The Review of Economics and Statistics 86: 195-210.

Eisenkopf, Gerald, Zohal Hessami, Urs Fischbacher, and Heinrich Ursprung. 2015. “Academic Performance and Single-sex schooling: Evidence from a natural experiment in Switzerland.” Journal of Economic Behavior & Organization 115: 123-143.

Eliot, Lise. 2013. “Single-Sex Education and the Brain.” The University of Chicago Medical School. Sex Roles 69: 363-381.

Fryer, Roland, and Steven Levitt. 2009. “An Empirical Analysis of the Gender Gap in Mathematics.” Working Paper 15430. Cambridge: National Bureau of Economic Research.

Gneezy, Uri, Kenneth Leonard, and John List. 2009. “Gender Differences in Competition: Evidence from a Matrilineal and a Patriarchal Society.” Econometrica Vol. 77: 1637-1664.

Goldin, Claudia, and Lawrence Katz. 2010. “Putting the Co in Education: Timing, Reasons, and Consequences of College Coeducation from 1835 to the Present.” Working Paper 16281. Cambridge: National Bureau of Economic Research.

Goodkind, Sara. 2012. “Single-sex Public Education for Low-Income Youth of Color: A Critical Theoretical Review.” Sex Roles 69: 393-402.

Halpern, Diane, et. al. 2011. “The Pseudoscience of Single-Sex Schooling.” Science 333: 1706-1707.

Hayes, Amy, Erin Pahkle, and Rebecca Bigler. 2011. “The Efficiacy of Single-Sex Education: Testing for Selection and Peer Quality Effects.” Sex Roles 65: 693-703.

Hoxby, Caroline. 2000. “Peer Effects in the Classroom: Learning from Gender and Race Variation.” Working Paper 7867. Cambridge: National Bureau of Economic Research.

Hughes, Teresa. 2007. “The Advantages of Single-Sex Education.” National Forum of Educational Administration and Supervision 23: 5-15.

Jackson, C. Kirabo. 2011. “Single-Sex Schools, Student Achievement, and Course Selection: Evidence from Rule-Based Student Assignments in Trinidad and Tobago.” Working Paper 16817. Cambridge: National Bureau of Economic Research.

Jimenez, Emmanuel, and Marlaine Lockheed. 1989. “Enhancing Girls’ Learning Through Single-Sex Education: Evidence and a Policy Conundrum.” Educational Evaluation and Policy Analysis 11: 117-142.

Lavy, Victor, and Analía Schlosser. 2011. “Mechanisms and Impacts of Gender Peer Effects at School.” American Economic Journal: Applied Economics 3: 1-33.

Lee, Soohyung, et. al. 2014. “All or Nothing? The Impact of School and Classroom Gender Composition on Effort and Academic Achievement.” Working Paper 20722. Cambridge: National Bureau of Economic Research.

Lee, Soohyung, Muriel Niederle, and Namwook Kang. 2014. “Do Single-sex schools make girls more competitive?” Economics Letters 124: 474-477.

Mael, Fred, et. al. 2005. “Single-Sex Versus Coeducational Schooling: A Systematic Review.” U.S. Department of Education. Doc #2005-01.

Pahkle, Erin, Janet Shibley Hyde, and Carlie M. Alison. 2014. “The Effects of Single-Sex Compared with Coeducational Schooling on Students’ Performance and Attitudes: A Meta-Analysis.” American Psychological Review 140: 1042-1072.

Park, Hyunjoon, Jere Behrman, and Jaesung Choi. 2013. “Causal Effects of Single-Sex Schools on College Entrance Exams and College Attendance: Random Assignment in Seoul High Schools.” Demography 50: 447-469.

Park, Hyunjoon, Jere Behrman, and Jaesung Choi. 2012. “Do Single-Sex Schools Enhance Students’ STEM (Science, Technology, Engineering, and Mathematics) Outcomes?” Working Paper 12-038. Philadelphia: Penn Institute of Economic Research.

Schneeweis, Nicole, and Martina Zweimuller. 2011. “Girls, Girls, Girls: Gender Composition and Female School Choice.” Economics of Education Review 31: 482-500.

Steele, Claude, Steven Spencer, and Joshua Aronson. 2002. “Contending with Group Image: The Psychology of Stereotype and Social Identity Threat.” Advances in Experimental Social Psychology 34: 379-440.

Yalcinkaya, M. Talha, and Alyse Ulu. 2012. “Differences between single-sex schools and coeducational schools.” Procedia – Social and Behavioral Sciences 46: 13-16.