A Female Advantage in Academic Hiring?
In their recent Proceedings of the National Academy of Sciences article about tenure-track hires, Williams and Ceci say:
A number of audits of hiring by universities have been reported in the past two decades and these have reported either a neutral playing field in non-mathematical fields . . . , or, more commonly, a pronounced female hiring advantage in math-intensive fields. . . . Here is what we know about the female advantage in real-world hiring of tenure-track applicants in STEM fields in the United States and Canada: There is a female advantage in all large-scale studies dating back to the 1980s. (SI Appendix, p. 26, emphasis added)
Williams and Ceci quantify “female preference” as the ratio of female hires to female applicants. However, they do not compute these ratios for the “audit studies” they cite. This post makes some of those computations.
Interestingly, three of the eight studies cited come from Canada, but some large-scale audit studies of United States universities are not mentioned. This post examines a few of the studies that could have been cited, finding that the claim above is not supported and offering an alternative explanation for the statistics in the US audit studies.
Update (August 2015): Ceci makes an assumption explicit (see comments section here and further discussion below).
Update (July 2015): The three Canadian studies concern hiring in the 1980s and 1990s. During this period, the gender distribution of Canadian faculty hires may have been affected by a “brain drain” to the United States. Studies note that PhDs were overrepresented among these migrants. What is known suggests that a very large proportion of these PhDs were male and among “the best and the brightest.” Thus, faculty hiring patterns in Canada and the United States may differ. Moreover, an exodus of men may be a factor in the apparent overrepresentation of women as hires in Canada. Further discussion is below.
Figure from Iqbal, M. (2000). Brain drain: Empirical evidence of emigration of Canadian professionals to the United States. Canadian Tax Journal, 48(3), 674–688.
Audit study 1 (cited in the PNAS article). Andrew Irvine (1996) examined the distribution of faculty members in Canada at selected years before 1988. In examining numbers of women by rank (but not field), he found that more women were hired than would be expected from estimated numbers of PhDs granted in Canada. He compared percentages of women in various faculty positions with percentages of doctorates granted to women in Canada, disaggregating them by field.
In Irvine’s study, the percentage of female applicants is measured by percentage of PhDs who are female, and percentage of female hires is measured by percentage of professors of various ranks who are female. (Thus, “applicants” are not necessarily people who apply. This point will recur in later examples.) I have put his statistics for assistant professors in the table below and computed the quotient of female assistant professors and PhDs (that is column 3 divided by column 2). Fields are listed in descending order, according to these quotients. Quotients greater than 1 appear above the dashed line. If these are interpreted as a “female hiring advantage,” then we see that the greatest female hiring advantage was in education, and the greatest disadvantage was in mathematics and physical sciences. This is just the opposite of “a neutral playing field in non-mathematical fields” and “a pronounced female hiring advantage in math-intensive fields.”
Update (August 2015): Ceci explains (see comments section here) that “It is extremely likely that the percentage of actual applicants for the mathematics-physical science professorships in Irvine’s analysis was significantly and considerably lower than 15%.” As I say in my reply, I do not think this assumption is warranted. Studies that record percentages of actual applicants are rare and most are confined to Research I universities. These studies do indeed find that percentages of actual applicants who are women tend to be lower than percentages of PhD recipients who are women. However, faculty demographics and the application patterns shown in the University of California audit (see Tables 5 and 10) suggest that percentages of women applying for tenure-track positions differ between high- and low-status institutions.
Update (July 2015): As noted above, significant numbers of Canadian PhDs move to the United States. Thus, using percentages of PhDs granted by Canadian universities (as done in Irvine’s study) may be a misleading measure of the gender distribution of applicants, as discussed further below.
Two other Canadian audit studies report on faculty hires during the 1990s. Seligman (2001) reports on hires at the University of Western Ontario. Kimura (2002) reports on hiring at half of the departments at Simon Fraser University and the University of British Columbia, finding that 71.1% of applicants were male. Of the 105 hired, 59% were men. Kimura remarks, “Assuming equivalent quality ranges in men and women applicants, . . . it must follow that, when preferences are severe, some women will be hired over better-qualified men” (emphasis added).
However, this assumption may not be warranted.
A Statistics Canada study (Frank & Bélair, 1999) found that 12% of people who earned doctorates from Canadian universities in 1995 moved from Canada to the United States. Of these (an estimated 359), 76% were men.
Among all “the class of 1995” degree recipients who left Canada for the United States:
- 93% in engineering and applied sciences were men.
- 72% in mathematics and physical sciences were men.
- 44% reported their rank in class as among the top 10%. (pp. 7, 38)
A later study for Human Resources and Skills Development Canada (Boudarbat & Connolly, 2013) suggests that these trends continue: PhDs are the most likely to leave, men are more likely to leave than women, and “mathematics, computer and information sciences and architecture, engineering and related technologies consistently stand out as fields for which there is a strong likelihood of moving to the United States” (p. 3).
Audit study 2 (not cited in the PNAS article). At a hearing about faculty hiring in the nine-campus University of California system in 2001, Chancellor M. R. C. Greenwood stated:
I must be honest and tell you that I have no rational explanation for the data on UC’s mathematics hires.
- In mathematics, UC hired 5.4% female faculty, the national Ph.D. pool is 22.1%, UC graduates 18.4% female Ph.D.s, our comparable institutions graduate 19.6% women Ph.D.s and even the postdoctoral pool (which in mathematics is highly competitive and is an indication of potential for future funding) has 13.2% females in it.
Williams and Ceci classify mathematics as “math-intensive.” However, the statistics above do not show “a pronounced female hiring advantage.”
Associated with the hearing is a detailed 115-page report: University of California: Some Campuses and Academic Departments Need to Take Additional Steps to Resolve Gender Disparities Among Professors, Bureau of State Audits, Sacramento, California. Its Table 12 shows percentages of female PhDs and professors by field.
The table below shows these percentages for fields that include those in Irvine’s categories. Like Irvine’s statistics, some entries in the table support “a neutral playing field in non-mathematical fields” and “a pronounced female hiring advantage in math-intensive fields”—and some don’t. However, the fields with “female advantage” are not all the same. For example, Irvine’s statistics show a “female advantage” in education and agriculture & biological sciences, but the UC statistics show the opposite.
Update: Table 5 of the UC audit report gives information about applicant gender for selected disciplines for each of the nine UC campuses. For mathematics, proportions of women in the applicant pools varied between 9% and 19%.
Here is the list for the nine campuses: 9%, 18%, 9%, 13%, 19%, 10%, 15%, 18%.
As shown below, 8% of assistant professors hired in mathematics were female.
Like Irvine’s statistics, some of the statistics in this table do support “a neutral playing field in non-mathematical fields” and “a pronounced female hiring advantage in math-intensive fields”—and some do not. Like Irvine’s statistics, the UC statistics for engineering and computer science suggest “a pronounced female hiring advantage.” The next study suggests this advantage may be illusory.
Audit study 3 (not cited in the PNAS article). Table 3 of a survey report from the Computing Research Association (CRA) shows the percent of newly hired tenure-track faculty in US and Canadian computing science and engineering (CS/CE) departments who are female by year. For 2007–08, that was 24% although, as shown in the survey’s Table 1, 19% of doctorates were female.
Does this indicate the “female hiring advantage” for computer science described by Williams and Ceci? It’s not so clear if we consider non-academic jobs. Surveys for 1985–2001 show that a hefty proportion of new PhDs in CS/CE go to industry, over half between 1994 and 2001 as shown in the figure below. A recent Science article notes this was also the case in 2013.
CRA’s report Recruitment and Retention of Faculty in Computer Science and Engineering (which was published in 2003) says:
The appeal of industry was stated by 47% of chairs as one of the main reasons why they lost faculty during the past three years. Salary and the presence of colleagues in the same research area were the most frequent industry attractions, according to department chairs.
The report’s survey of CS/CE job-changers suggests that salary tends to be a more important factor for men than women: 43% of men selected “salary” as a reason for leaving vs 13% of women (p. 19). It may be that disproportionately more men than women applied for jobs in industry. Thus, we don’t know if the 24% female assistant professors vs 19% female doctorates is due to “female hiring advantage,” to disproportionately fewer men applying for academic jobs, or to stronger female applicants.
This report also notes that in BA-granting departments, women were 32% of new hires, but 22% of those in PhD-granting departments. (A similar phenomenon occurs for mathematics, see the table in my previous post.)
The report also offers some interesting ideas about female applicants, including a possible explanation for women’s higher ratio of offers to applications.
As new PhDs, women submitted far fewer applications than men and received many more offers per application. Female new hires applied for only 6 positions (compared with 25 for men), obtained 0.77 interviews per application (vs. 0.37 for men), and received 0.55 offers per application (vs. 0.19 for men). Obviously women were much more selective in where they applied, and also much more successful in the application process, but the reasons for this selectivity are unclear:
- Women may have been more efficient in their searches, only applying to the positions that fit them best, or they may have underestimated their value and submitted fewer applications than they should have.
- The latter theory is partly supported by two additional results from our survey. First, the job-changers’ survey shows women applying for slightly more positions than men. This may indicate that women have learned how to game the system more effectively or have gained more confidence in their abilities. Second, if women are underestimating their value, they may be applying to schools with lower hiring standards, which would explain their high ratio of applications to offers. (p. 31)
Audit study 4 (not cited in the PNAS article). Yet another study that Williams and Ceci might have examined is Donna Nelson’s diversity survey. The table below is excerpted from “Statistical Trends in Women’s Participation in Science: Commentary on Valla and Ceci (2011),” which drew on Nelson’s survey findings. Note that, in contrast to the statistics above, computer science does not show a “female advantage.”
Audit study 5 (cited in the PNAS article). An article by Deborah Merritt and Barbara Reskin examines tenure-track law school hires of recent graduates between 1986 and 1991. An interesting feature of some law schools is that recent graduates may be hired at different ranks, as assistant or associate professors, both tenure track. Women were slightly more likely to get tenure-track jobs at more prestigious universities, but less likely to get higher ranked tenure-track jobs. Is that “a neutral playing field”? I don’t think so, but it appears that Williams and Ceci do.
My conclusions. Williams and Ceci claim there is “a neutral playing field in non-mathematical fields” and “a pronounced female hiring advantage in math-intensive fields.” If “neutral playing field” is operationalized as “quotient of female assistant professors to female PhDs is 1” or as “women and men obtain positions of equivalent rank,” the studies discussed above do not support their claim.
These studies illustrate the shortcomings of Williams and Ceci’s classification of fields as “mathematically-intensive” or “non-mathematical,” and the pitfalls of considering only tenure-track positions. Depending on field, a recent PhD may also obtain a variety of other positions, e.g., in industry or government, as a postdoc, at a research institute, or as an adjunct. The distribution of PhDs among different types of jobs varies with economic conditions, e.g., dot-com boom or decreased state funding for higher education. It also varies by field, e.g., many (currently about 70%) statistics and biostatistics doctorates obtain jobs in government or industry, or at research institutes. As discussed in my recent Mathematical Intelligencer article, I think that the situation for women (and men) in STEM is better described in terms of occupational stratification and segregation. In some fields (e.g., law), the jobs with high pay and prestige (which tend to attract more men) tend to be outside of academe. In other fields (e.g., mathematics), the jobs with high pay and prestige tend to be within academe, at PhD-granting universities rather than BA- and MA-granting institutions.