Women considering careers in academic science confront stark portrayals of the treacherous journey to becoming professors. Well-publicized research depicts a thicket of obstacles standing between female graduate students and tenure-track positions, including inadequate mentoring and networking (1); a chilly social climate (2); downgrading of work products such as manuscripts (3), grant proposals (4), and lectures (5); and gender bias in interviewing and hiring (6–9). Numerous blue ribbon panels and national reports have concluded that implicit, and sometimes explicit, attitudes pervade the hiring process and negatively influence evaluations of female candidates and their scholarship, contributing to women’s underrepresentation within the academy (e.g., refs. 10–13).

Women’s underrepresentation in academic science is hardly trivial. In life and social sciences, women now earn the majority of doctorates, but they make up a minority of assistant professors. In 1993–1995, 28.4% of assistant professors were women, but 41.6% of Ph.D.s awarded in the same cohort went to women. That is, almost one-third of the women did not advance from receiving their Ph.D. to an assistant professorship (see ref. 14, figure 5). More recently, in 2008–2010, this gap widened to 22 percentage points (53.2% of doctorates to women; 31.6% of assistant professorships to women), and this gap persisted after controlling for demographics, degree characteristics, and field (15). [This winnowing of women in the STEM (science, technology, engineering, mathematics) tenure-track pipeline is a result of women Ph.D.s being far less likely than men to apply for tenure-track jobs, rather than to women applying but being rejected at higher rates than men (14).] Against this bleak backdrop, it is perhaps no surprise that talented young women opt out of the STEM tenure track either by not applying for assistant professorships at the same rate as men or, in some fields, by not majoring in them in college in the first place (14).

The point at which scientists choose to apply for tenure-track assistant professorships is a key juncture in understanding the problem of women’s underrepresentation. Once hired, women prosper in the STEM professoriate (14, 16–18): They are remunerated, persist, and are promoted at rates roughly comparable to men’s (14) after controlling for observable characteristics, including academic productivity. However, to be hired and eventually tenured, women must first apply. Unfortunately, despite their success once hired, women apply for tenure-track positions in far smaller percentages than their male graduate student counterparts (14, 16, 18). Why might this be?

One reason may be omnipresent discouraging messages about sexism in hiring, but does current evidence support such messages? Despite this question’s centrality to any informed discussion about women’s underrepresentation in academic science, only one experimental study (7) contrasted faculty ratings of the relative “hirability” of hypothetical identically qualified women and men. Results showed that both female and male psychology faculty members downgraded a hypothetical woman’s academic record compared with an identical man’s. However, this study was published 16 y ago and involved only one field, psychology, a discipline that is more than 50% female (14).

There are two critical omissions from the current data landscape. First, no experimental study of tenure-track faculty hiring in math-intensive fields has ever evaluated whether bias can explain women’s underrepresentation in those fields today. This is important because it is in math-intensive fields that women are most underrepresented (see ref. 14 for historical and current data). Second, no current experimental study demonstrating sexist hiring is based on actual faculty members who were asked to rate identically qualified tenure-track assistant professor applicants. Instead, past studies have used ratings of students’ hirability for a range of posts that do not include tenure-track jobs, such as managing laboratories or performing math assignments for a company. However, hiring tenure-track faculty differs from hiring lower-level staff: it entails selecting among highly accomplished candidates, all of whom have completed Ph.D.s and amassed publications and strong letters of support. Hiring bias may occur when applicants’ records are ambiguous, as was true in studies of hiring bias for lower-level staff posts, but such bias may not occur when records are clearly strong, as is the case with tenure-track hiring. Thus, we focused on male and female assistant professor candidates who were identically and unambiguously strong and used tenured/tenure-track STEM faculty members from all 50 US states as raters to determine the role of gender bias in tenure-track faculty hiring.

This program of research consisted of five experiments involving 873 tenure-track faculty members from 371 colleges and universities, spanning 50 states and the District of Columbia. We investigated faculty hiring preferences for hypothetical applicants in two math-intensive fields in which women are substantially underrepresented (engineering and economics) and two non–math-intensive fields in which women are well represented (biology and psychology; see ref. 14, figure 4A, for current and historical data on women in these fields). We used (with embellishments) a method for revealing gender bias in hiring that has been used frequently in past studies (6, 7, 19): We compared the likelihood of identically qualified women and men (named Dr. X, Dr. Y., and Dr. Z, and differing across faculty raters solely in the gender pronoun used to refer to them) being ranked first by individual faculty members for a tenure-track assistant professorship (SI Appendix).

To illuminate contextual factors in faculty decision-making, we also studied the effects of candidate lifestyle on hiring preference for otherwise-identical male and female candidates. For example, in experiment 1, we included two applicants, both of whom were unmarried and childless or, in another condition, both of whom were married with preschool-age children and stay-at-home spouses, to see how a candidate’s perceived hirability was influenced by various lifestyle situations. In experiment 2, which focused on nonmatching lifestyles, we included a married father of two with a stay-at-home spouse competing against a single mother of two with an absent ex-spouse (in addition to a third, “foil,” candidate appearing in every contest, as described later). Because so much research points to fertility decisions as key determinants of women’s decisions to opt out (20–23), in experiment 3 we explored the relative hirability of identical mothers (or in a separate condition, identical fathers) who either took or did not take a 1-y parental leave in graduate school. [Women’s perceptions that an extended maternity leave will cause them to be viewed as less committed to their profession (22, 23) may influence some women to opt out entirely.] Experiment 4 used the same method as experiment 1, except faculty members rated candidates’ full curricula vitae (CVs), rather than narrative summaries. In experiment 5, faculty members rated only a single applicant (female or male, with identical records) to ensure the findings of experiments 1–4 were not a result of socially desirable responding.

Our research required a design enabling a realistic comparison of a woman and man with identical credentials applying for the same assistant professorship without signaling our key hypotheses to faculty. We could not simply send faculty members two identical candidate descriptions differing only in gender and ask which person the faculty member preferred to hire. Such a transparent approach would have revealed our central question and compromised the results. Thus, the design of experiments 1–4 was necessarily complex, relying on comparisons among faculty members to reveal how hiring preferences were influenced by candidate gender and lifestyle. Faculty members received information regarding three short-listed candidates for a tenure-track assistant professorship in their department. This information included the search committee chair’s notes reflecting the committee’s evaluation of each candidate’s scholarly record (based on having read the vita and actual publications); the information also included excerpts from letters of reference and the faculty’s average ratings of the candidate’s job talk and interview. Narrative summaries of candidate credentials rather than CVs were used because of our sampling over four fields and three major Carnegie-institution levels while needing to simultaneously hold constant applicant credentials (i.e., no single vita would be realistic across both large, research-intensive universities and small colleges). In addition, the chair’s comments about fit with the department were noted: “Dr. Z: Z struck the search committee as a real powerhouse. Based on her vita, letters of recommendation, and their own reading of her work, the committee rated Z’s research record as ‘extremely strong.’ Z’s recommenders all especially noted her high productivity, impressive analytical ability, independence, ambition, and competitive skills, with comments like, ‘Z produces high-quality research and always stands up under pressure, often working on multiple projects at a time.’ They described her tendency to ‘tirelessly and single-mindedly work long hours on research, as though she is on a mission to build an impressive portfolio of work.’ She also won a dissertation award in her final year of graduate school. Z’s faculty job talk/interview score was 9.5/10. At dinner with the committee, she impressed everyone as being a confident and professional individual with a great deal to offer the department. During our private meeting, Z was enthusiastic about our department, and there did not appear to be any obstacles if we decided to offer her the job. She said her husband is a physician (with opportunities to practice in our area) and that they will need two slots at university daycare for their children. Z said our department has all of the resources needed for her research.” (In the full-vita condition, experiment 4, faculty members evaluated actual full CVs of engineering applicants in place of summaries.) We asked faculty members to rank the candidates first, second, and third for the job.

Twenty counterbalanced sets of materials for experiments 1–3 were developed in which the same candidate was depicted as “she” in one version and “he” in the next. We also varied lifestyle for candidates in the same way, using identical candidate records of accomplishments but varying descriptors such as “single with no children,” “married, needs two slots in university daycare,” or “divorced, needs two slots in university daycare”: disclosures portrayed in our materials as having been voiced spontaneously by candidates themselves, as sometimes happens during job interviews. One tactic to disguise the hypothesis and increase the realism of the rating task was the use of a pretested foil candidate in each set of three candidates. The foil candidate was created to be slightly weaker than the other two candidates but still excellent. The foil rounded out a slate of three candidates and allowed us (in certain experimental conditions) to present two male and one female candidate to faculty, which was important in creating a realistic short-list of job finalists, especially in disciplines such as engineering and economics, which have far fewer women than men applicants. Another disguise tactic was the incorporation of ecologically rich personal descriptions of candidates that included their demeanor at the search committee dinner. Two forms of these descriptions were developed: one with traditionally female adjectives (“imaginative, highly creative, likeable, kind, socially skilled”) and one with traditionally male adjectives (“analytical, ambitious, independent, stands up under pressure, powerhouse”) (24, 25). In every case, actual candidate gender was counterbalanced with adjective gender across versions, so that half of the faculty members were sent a given candidate depicted as a woman described with male adjectives and half were sent the same woman depicted with female adjectives, and vice versa for male candidates.

In all 20 sets of materials in experiments 1–3, the contest was between Drs. X, Y, and Z (with Y being the foil candidate). Faculty members were sent emails containing one of the 20 versions and asked to rank the three candidates for an assistant professorship in their department. Versions were sent out randomly after stratification of the sample by the three major Carnegie classifications of institutions (doctoral-granting universities, master's colleges and universities, and baccalaureate colleges), department (biology, engineering, economics, psychology), and faculty gender. The 20 versions of materials allowed us to compare rankings between faculty raters of Dr. X as a woman with rankings of Dr. X as a man. Again, both had identical credentials and were described with identical adjectives; the sole difference was the gender pronoun used to describe Dr. X: “he” in one version and “she” in the other. The same manipulation was performed for Dr. Z. (Remember that Dr. Y was the foil, pretested to be slightly weaker than Dr. X and Dr. Z, but still excellent; the foil was ranked first 2.53% of the time.)

Our method enabled us to contrast the hirability of women and men candidates in a between-subjects design, with counterbalancing of candidate gender and adjective gender between faculty members (half female, half male), so that rankings of identically qualified candidates of different genders and lifestyles could be compared across 873 faculty from 371 institutions in 50 US states. Multiple validations (described here and in the SI Appendix) were incorporated into our design, including use of full vitas to ensure results matched those for vita summaries, use of a paid sample with 91% response rate to validate results from the national sample, and validation of adjectives to ensure past findings of gender connotations applied to faculty in our study. Our experimental design thus disentangled the relative effects on hiring preferences of applicant gender, applicant lifestyle, faculty gender, and faculty field.