For centuries men dominated academic science and engineering. Gender bias once greatly imperiled the progress of any woman inclined to pursue science, technology, engineering or math (STEM). Over the past 40 years, however, society has gradually begun to accept, if not embrace, the notion of the female biologist, mathematician or engineer, and the number of women in science at all levels has increased dramatically. In the early 1970s women received 29 percent of bachelor's degrees and 10 percent of Ph.D.s in STEM. By 2011 women held just over half of STEM bachelor's degrees, about the same as their proportion of high school graduates, and 41 percent of Ph.D.s.

Yet those promising statistics belie two areas of stagnation for women in STEM. First, a marked deficit of women remains in certain disciplines: geoscience, engineering, economics, mathematics, computer science and the physical sciences. Although women outnumber men among college graduates overall, they currently make up only 25 percent of college majors in these math-heavy fields, and their numbers have been dropping since 2002. Once a woman enters one of these fields, however, her progress toward a Ph.D. and a tenure-track job resembles that of her male counterparts. About 10 percent of both men and women with college degrees in math-intensive fields proceed to Ph.D.s, and 35 to 38 percent of those with Ph.D.s receive assistant professor jobs. According to the U.S. National Science Foundation's 2010 Survey of Doctorate Recipients, women make up 30 percent or less of assistant professors and 7 to 18 percent of full professors in these disciplines.

Meanwhile, in psychology, life sciences and social sciences, the reverse situation persists: women's ranks among college majors have overtaken those of men, but the female presence shrinks at higher levels. The percentage of female assistant professors in psychology, life sciences and social sciences from 2008 to 2010 was 31.6 percent, compared with 53.2 percent in the corresponding Ph.D. years—a gap of 22 percentage points. Women now hold just 40 percent of full professorships in psychology and about a quarter of these in life and social sciences.

People have put forth myriad explanations for these phenomena—from biased interviewing, hiring and promotion to a gender disparity in quantitative ability and career preferences. Our research group has whittled away at these explanations to determine the most influential ones for each gap.

For the math-centered academic fields, in which women are scarce by the end of college, a difference in interests is a more important factor than a disparity in ability. Cultural messages suggest (incorrectly) that girls are not competent in math. That misconception, along with a lack of female mentors, may influence the number of girls seriously considering these fields. Later on, women who pursue biology, psychology and other social sciences are less likely to get stuck on the ladder—say, because of institutional bias—than to simply jump off, opting out of the competition for faculty positions because they do not believe these jobs are compatible with having families. Addressing the tendency of women to leave academic science will mean making these fields more amenable to work-life balance.

The Math Gap

The percentage of women receiving advanced degrees in a field is negatively correlated with that area's mathematical content. Simply put, the more math, the fewer women. Many have suggested that differences in mathematical aptitude might contribute to this gender gap.

Girls and boys do not differ in average mathematical ability. Using national probability samples involving millions of school-aged children, psychologist Janet S. Hyde of the University of Wisconsin–Madison and her colleagues have repeatedly found that the average performance of boys on standardized math tests almost entirely overlaps with that of girls. The researchers' 1990 meta-analysis of 100 studies involving three million children revealed no sex differences at any age or for problems of any complexity, except that boys performed slightly better on advanced high school–level math problems. By the beginning of the 21st century, however, girls had evened the score on the hardest problems on the National Assessment of Educational Progress for high school students—most likely a result of having taken more math courses.

Of course, the individuals entering math-intensive fields are not average in mathematical ability, and at the high end of the math scale boys do outnumber girls. For example, in a study published in 2009 educational psychologists David F. Lohman of the University of Iowa and Joni M. Lakin, now at Auburn University, analyzed 318,599 American third to 11th graders and found that among the top 4 percent of high scorers, two thirds of the kids were male and one third female. Similarly, in 2013 psychologists Gijsbert Stoet, now at the University of Glasgow, and David C. Geary of the University of Missouri found slightly more than twice as many boys as girls in the top 1 percent of scores from 15-year-olds on the Program for International Student Assessment from 2000 to 2009.

A 2-to-1 ratio favoring male students could be instrumental in impeding women from achieving at high levels in math-intensive fields. Yet the average quantitative scores of Ph.D. candidates in the most math-heavy fields hover around the 75th percentile, a level where the sex gap is considerably less than 2 to 1. For that matter, sex differences in mathematics scores do not translate into grades in math classes (in which female students fare slightly better), and 40 to 48 percent of baccalaureates in mathematics have been awarded to women for four decades. (Although women are majoring in math in numbers approaching those of men, fewer women are majoring in most of the math-related fields.)

What is more, this disparity in test scores does not exist everywhere or in every ethnic group, suggesting it is mutable. In the U.K., the male-to-female ratio at the top 4 percent is smaller: 3 to 2 compared with 2 to 1 in the U.S. In Iceland, Singapore and Indonesia, more girls than boys scored in the top 1 percent at certain ages. In 2009 Hyde and oncologist Janet E. Mertz of the University of Wisconsin–Madison School of Medicine and Public Health found a similar pattern among Asian-Americans, with more female students appearing in this high-scoring group. Thus, sociocultural factors are driving some of the sex differences among the young math elite.

The Culture Gap

Among these influences is a phenomenon known as stereotype threat. Researchers such as social psychologist Claude M. Steele, now at the University of California, Berkeley, have shown that the awareness that others expect members of a social group to do poorly in math is sufficient to create anxiety and worse performance among members of that group. Even subtle priming of sex can hamper female performance in math. For example, female test takers who marked the gender box after completing the SAT Advanced Calculus test scored higher than female peers who checked the gender box before starting the test (although some critics of this work have claimed that the effect is not as robust or as large as claimed).

Some of the attitudes that lead to stereotype threat also may depress girls' interest in math-intensive fields. In a 2012 review of recent work sociologist Lara Perez-Felkner of Florida State University and her colleagues revealed that by age five, girls receive the message that math is for boys. By middle school, 9.5 percent of boys expect to work in science or engineering compared with less than half as many girls (4.1 percent).

These early preferences often change in individuals. But in high school, when they are far more stable, a gender tilt remains. In a book published in 2003 sociologists Yu Xie of the University of Michigan and Kimberlee A. Shauman of the University of California, Davis, found that among college-bound high school seniors, more than three times as many males as females expected to major in science and engineering. In 2013 sociologist Stephen L. Morgan of Cornell University and his colleagues reported that male high school students were more than four times as likely as female students to have listed only STEM occupations apart from the life sciences (and excluding medical, biological, health and clinical sciences) in their plans.

These findings are in line with data showing that females tend to prefer people-oriented professions such as nursing, counseling and teaching. (For the ratio of boys to girls taking science AP exams, see the upper left illustration on the next page.) Overall, studies show that high school students' expectations of their future college major explain 28.1 percent of the gender gap in science and engineering baccalaureates.

College experiences are also important for recruiting women into math-intensive fields. Women who major in science and engineering are more likely than men to start down that path in college rather than in high school. In addition to exposure to these fields, women need role models. According to a 2010 study led by economist Scott E. Carrell of U.C. Davis, female students at the Air Force Academy who had female professors in introductory STEM courses (which are randomly assigned at the academy) were more likely to pursue a STEM major than were peers assigned to male professors.

Women may need more academic support to stick with a science major because they attach greater importance to getting high grades than men do and are therefore more likely to drop courses in which their grades may be lower—the so-called fear of a B. In a 1997 study sociologist Elaine Seymour of the University of Colorado Boulder and historian Nancy Hewitt of Rutgers University found that loss of self-esteem caused by low grades in introductory science and math courses was associated with women's leaving science and engineering majors.

Challenging Bias

The women who do graduate with majors in math-heavy fields, however, advance in them just as often as men do. Women held 26.3 percent of the Ph.D.s in these fields in 2011, a figure that mirrors their numbers among baccalaureates in the same fields seven years earlier, up from 16.8 percent in 1994; their proportion of assistant professorships rose from 14.3 to 22.7 percent, percentages not significantly different from those of women obtaining Ph.D.s in these fields five to six years earlier.

In contrast, in psychology and in the life and social sciences, the attrition of women after college is considerable. Since the mid-1990s women have received 60 percent or more of the bachelor's degrees in these fields; in 2011 they received 57.9 percent of the Ph.D.s, up from 46.1 percent in 1994. Yet although women still hold the most assistant professorships in psychology, men significantly outnumber them as assistant professors in the life sciences.

Many people believe that gender bias remains a signifi-cant factor in this drop-off (although why the discrimination would affect the life and social sciences more than mathematical fields is not clear). Numerous small experiments involving hypothetical applicants point to the existence of gender bias. For instance, in 2012 psychologist Corinne Moss-Racusin, now at Skidmore College, and her colleagues reported that when 127 science faculty at six U.S. universities evaluated fictitious applicants with bachelor's degrees for a laboratory manager post, they rated males higher and recommended higher starting salaries and more mentoring for them than for female applicants, even though there was no difference between their applications.

Yet real hiring data are inconsistent with the results from these mock situations. A 2010 U.S. National Research Council survey looked at hiring in six STEM disciplines—biology plus five math-intensive fields—including almost 500 departments at 89 prestigious universities nationwide from 1995 through 2003. The women who applied for tenure-track assistant professor positions were invited to interview and offered positions in each of these fields more often than would be predicted by their fraction of the applicant pool. Women were also offered posts for more senior, tenured positions at rates higher than their fraction in the applicant pool. Two other large-scale analyses, conducted in 2008 and 2010, similarly showed that women are hired at rates comparable to or better than men. What is more, no counterevidence exists. Simply put, no real-world hiring data show a bias against hiring women.

Of course, bias could still be a factor if female applicants are, on average, better than their male competitors. In that case, the high proportion of female Ph.D.s hired might mask bias that prevented an even greater percentage of women from getting jobs. There is no evidence of such superiority among female candidates, however. In a study submitted for publication in 2014 two of us (Ceci and Williams) sampled faculty from 347 universities and colleges in the U.S. to look for bias in the hiring of tenure-track assistant professors in STEM fields and failed to show that female superiority in hiring outcomes was related to objectively higher-quality female applicants. Moreover, objective measures of productivity such as publications do not indicate that women in the applicant pool are stronger than men.

Real-world data may conflict with those from fabricated hiring situations, in part because committees or departments, not individuals, typically make real academic hiring decisions. This process may mitigate the effects of bias, given people's reluctance to publicly express prejudice. In addition, many of the most prominent studies showing bias, including the one led by Moss-Racusin, did not involve academic tenure-track jobs but lower-level posts. Such studies also depicted candidates as having ambiguous academic credentials, which may create more room for bias than would occur when a candidate is clearly competent, as those in real academic hiring situations typically are.

Opting Out

Gender bias in science most likely still exists. But the data do not convince us that instances of bias meaningfully contribute to the growing deficit of female assistant professors in psychology and in the life and social sciences. Instead of being denied jobs in academic science, women are leaving on their own.

Currently the most important barrier to women advancing in STEM fields, at least in statistical terms, is the perception among female Ph.D. recipients and postdoctoral fellows that academic positions are not compatible with families. In work published in 2011 legal scholar Mary Ann Mason of the U.C. Berkeley, and her colleagues found that women Ph.D.s with no children and no plans to have children fared as well as men in applying for and getting STEM tenure-track jobs. In contrast, those who planned to have children opted out of the tenure-track pipeline at research universities in favor of careers they believed were more compatible with their plans, such as positions at teaching-intensive colleges or adjunct posts. In the survey, 28 percent of female postdocs who planned to start families opted out compared with 17 percent of men who anticipated having children; for those who already had children before taking their postdoc position, the attrition rates were 32 percent for women and 19 percent for men.

In 2007 pharmacologist Elisabeth Martinez of the University of Texas Southwestern Medical Center and her colleagues reported similar child-related attrition in a survey of 1,300 postdocs at the U.S. National Institutes of Health. Concerns about children may have a particularly strong impact on obtaining a tenure-track job in the life sciences because postdoc positions in these fields require long hours with little discretion over when those hours are. In addition, women in the life sciences have abundant opportunities to do science in nonacademic settings. As of 2008, 55 percent of jobs for biomedical Ph.D.s were outside of academia. Reviewing 2010 statistics from the Survey of Doctorate Recipients, two of us (Ginther and Kahn) found that women with Ph.D.s in psychology and the life and social sciences were the ones least apt to pursue tenure-track positions and most likely to have shorter hours in their nonacademic jobs than professors do.

Women also exit the labor force more than men do. Analysis of the 2010 doctorate survey indicates that women were more than twice as likely as men to have left the labor force—either from a job or right after getting a Ph.D. Most of those who were not retiring stepped out because of family considerations, compared with only a third of the few nonretiring men. Combining those who move from an academic post to another job with those who are no longer employed or who are seeking a job (but are not retired), women are indeed more likely than men to leave science: 12.5 versus 9.6 percent.

Promotion

Women are less likely than their male colleagues to be full professors in STEM disciplines, a 2012 study from the National Science Foundation found. Although there are no significant sex differences in promotion to tenure and full professorships in most math-intensive fields, women are promoted significantly less often in the life sciences, psychology and economics, according to two analyses of the doctorate recipient survey.

One probable contributor to this gap is that, as assistant professors, women are less productive than men. The 2008 doctorate recipient survey indicates that the average difference in their five-year publication count—the number of articles accepted by peer-reviewed journals—is 2.1 fewer articles for female academics, which represents a productivity gap of 19.6 percent. The reasons for this gap are unclear. Women do not seem to work less than men: the 2010 Survey of Doctorate Recipients shows surprisingly small gender differences in the weekly hours that tenure-track STEM faculty worked outside the home.

Children are not a primary cause, either. Using data from the 2008 Survey of Doctorate Recipients, we found that women without children publish noticeably more than women with children only in geoscience and psychology. Thus, except for these two fields, the presence of children cannot explain the gender-specific difference in publications. (Other surveys suggest that having children does reduce academics' work hours, but this effect is similar for both genders.)

Higher demands on female faculty's time for teaching or service could be important, however. In their 2003 book Xie and Shauman found that faculty teaching 11 or more hours a week had much lower research productivity. Some data suggest that female faculty spend relatively more time teaching than male faculty do. In our 2010 analysis of tenure and tenure-track scientists in the Survey of Doctorate Recipients, 37 percent of men listed research as their primary work activity, whereas only 31.5 percent of women did; in contrast, 53 percent of women versus 47 percent of men listed teaching as their primary work activity.

Many female academics may enjoy being mentors and teachers, of course, a pattern that would be consistent with women's overall preference for people-oriented fields. Nevertheless, if we are serious about encouraging women to enter math-heavy fields and enabling them to advance in psychology and the life and social sciences, we need to deal with the most important factors holding them back.

The deficit of girls in math-intensive disciplines is apparent by college. For cultural or other reasons, girls lean toward fields that involve living things rather than objects, an inclination that is evident by middle school. From a young age, female students profess to be more interested in medicine, biology, law, psychology and veterinary medicine, whereas males gravitate toward engineering and computer science. Although different interests may be acceptable, if we want to encourage girls to pursue math-intensive areas of study, we should find ways to mentor and support high school and middle school girls in math and science. We should also urge all entering college students to take science and math as early as possible, given that women often switch into math-centered STEM fields after starting college. Female role models are important for recruiting girls and women into science fields and keeping them there.

The problem of retaining women in the pipeline after college graduation occurs mainly in psychology and in the life and social sciences. Our analysis shows the roots of that problem lie less in biased hiring decisions and more in the choices women make to maintain work-life balance. Striking such a balance is particularly difficult in academia. Even so, female faculty members can and do become mothers, and university policies such as stopping the tenure clock and paid parental leave are a step toward making the tenure track compatible with the lives that many women want.

More adjustments may be needed, however. For instance, an option of part-time tenure, in which women with academic jobs work reduced hours, might help boost the number of female faculty. And a culture that is more accepting of family commitments—one in which faculty meetings are scheduled around them, for instance—may be necessary to encourage more talented women to contribute to important advancements in science.