I have neither the time nor the space to sum up either of these two papers (the second is a short supplement to the first), but if you’re interested in gender parity in STEM fields, you should definitely read the longer Stewart-Williams and Halsey paper. It’s fairly new (2021) and is loaded with data and references about the widely-discussed deficit (“inequity”) of women in some STEM fields, what factors might cause it, and what, if anything, should be done to assure parity. It’s a big paper—26 pages of text—but also has nearly every reference up to 2021 that I know about on the topic (and many more), with over 11 pages of citations in addition to the text.
You can read the paper by clicking on the screenshot below, or downloading the pdf here (reference at the bottom of the page). Stewart-Williams is a professor of psychology at the University of Nottingham in Malaysia, while Lewis Halsey is a Professor of Environmental Physiology at the University of Roehampton.
There are many aspects of the paper, but the overall message is that a lack of equity between men and women in some (not all) STEM fields cannot be wholly imputed to bias or “structural sexism” because there are many other factors causing such inequities. These factors include sex-differing preferences, interests, the greater overall variability in performance (and other traits) of men, evolution, and so on. The authors do note that there is evidence for bias against women, describing a long list of studies, but also show that there’s also evidence that women are favored in entering and succeeding in STEM, giving an even longer list of studies. We all know—though few mention—that the proportion of women in STEM goes down as countries become more equal in opportunity afforded to males and females, which suggests that in more gender-equal countries women’s preferences and other non-biasing factors are more freely excercised, perhaps leading to a decline in participation in STEM (I’ve written about this before).
Stewart-Williams and Halsey attribute some of the sex differences in interests (and variability) to evolution, but freely admit that any hypotheses they have are just stories and are very hard to test.
The biological difference in STEM representation can, say the authors, be partly imputed to the claim that “Men are more interested in things than are women, who in turn are more interested in people.” (Remember, this is an average, and doesn’t imply anything about whether some women can be more interested than many men in STEM fields, nor does it buttress any discrimination.) There are many studies implying that such differences are not only cultural universals among many societies (of course, one could argue that this is forced onto women by sexism in all societies), but they are also seen in very young infants who haven’t yet had a chance to be “socialized in sexism”, as well as in our primate relatives. These two points make an explanation based wholly on socialization less likely.
Rather than go into more detail, I’ll just say that if you strive for equity in gender or sex in STEM because you think inequities result solely from bias or sexism, do read this paper first. I’ll give one figure, below, and reproduce conclusion of the article.
First, a simplified diagram from the paper showing the many sources of inequities in sex representation in STEM. Each is discussed in detail in the paper:
. . . and the paper’s conclusion. I’ve put part of it in bold because I agree with the goal of maximizing opportunity rather than enforcing pure equity and making unevidenced claims of bigotry.
Conclusion: Many factors at play
In summary, any exhaustive discussion of the relative dearth of women in certain STEM fields must take into account the burgeoning science of human sex differences. If we assume that men and women are psychologically indistinguishable, then any disparities between the sexes in STEM will be seen as evidence of discrimination, leading to the perception that STEM is highly discriminatory. Similarly, if we assume that such psychological sex differences as we find are due largely or solely to non-biological causes, then any STEM gender disparities will be seen as evidence of arbitrary and sexist cultural conditioning. In both cases, though, the assumptions are almost certainly false. A large body of research points to the following conclusions:
- that men and women differ, on average, in their occupational preferences, aptitudes and levels of within-sex variability;
- that these differences are not due solely to sociocultural causes but have a substantial inherited component as well; and
- that the differences, coupled with the demands of bearing and rearing children, are the main source of the gender disparities we find today in STEM. Discrimination appears to play a smaller role, and in some cases may favour women, rather than disfavouring them.
These conclusions have important implications for the way academics and policy makers handle gender gaps in STEM. Based on the foregoing discussion, we suggest that the approach that would be most conducive to maximizing individual happiness and autonomy would be to strive for equality of opportunity, but then to respect men and women’s decisions regarding their own lives and careers, even if this does not result in gender parity across all fields. Approaches that focus instead on equality of outcomes – including quotas and financial inducements – may exact a toll in terms of individual happiness. To the extent that these policies override people’s preferences, they effectively place the goal of equalizing the statistical properties of groups above the happiness and autonomy of the individuals within those groups. Some might derive different conclusions from the emerging understanding of human sex differences. Either way, though, it seems hard to deny that this understanding should be factored into the discussion.
People will of course bridle at the claim that there’s a “substantial inherited component” to gender disparities, crying that “it’s evolutionary psychology—Nazism!”. But there’s ample evidence that men and women differ in morphological and behavioral ways that can be explained (though not “proved”) by evolution. This of course goes against the “progressive” conclusion that men and women are on average identical in every trait except perhaps in those morphological differences (size, build, genitalia) connected with the biological basis of sex. But those who believe that men and women are identical in every aspect of thought, behavior, and mentation are fighting a wealth of data. (I have to emphasize again that differences do not imply superiority or inferiority, but that’s so obvious that I shouldn’t have to say it for the umpteenth time.)
**************
The paper below is basically a short gloss on the paper above, and provides more data supporting the claim that while sex inequities in STEM can (and do) result partly from bias against women, that bias “cannot explain the corpus of findings related to gender differences in math-intensive disciplines. Click the screenshot to read it, and you can find the pdf here (reference at the bottom of the page).
The authors did their own three-year analysis of gender bias in six areas (letters of recommendation, tt [tenure track] hiring, journal acceptances, grant funding, salary, and teaching ratings). The fields surveyed aren’t listed, as the study isn’t yet published, but they found one area in which there was gender bias: “students of both genders rate women instructors’ teaching skills lower than men” [sic]. This is an average and shows heterogeneity among areas.
They found possible gender bias in “the academic salary gap”, but qualify it a bit:
In the second domain in which there is a possible gender bias—the academic salary gap—the presence or absence of bias is less clear. Although we tilted toward a bias explanation, we were unable to make an airtight case for it. The average gender salary gap in academia writ large is around 18%, but much of this is explained by the type of institution (e.g., two-year and four-year colleges, large research-oriented universities), discipline (more women are employed in lower-paying humanities fields than in higher-paying engineering and business fields), and years of experience. With these as controls, the gender difference among those on tt is less than 4%. And that difference might be even smaller if studies are able to control for productivity (publications), which no study of the salary gap has done. The evidence on publications, which we also summarized in our paper, points to gender differences in publishing, so this could account for the remaining 4% salary gap. So we are agnostic. We concluded that the evidence might point to some bias in salaries—although it is much smaller than averages suggest—and might not be the result of gender bias.
Finally, they report “no systematic gender bias” in the other areas over a long period of time:
In the other four domains (letters of recommendation, tt hiring, grant funding, and journal success) we came to the conclusion that there was no systematic gender bias in the last 15–20 years. Looking at studies that directly measured tt outcomes such as the likelihood of grant application success, acceptance of journal submissions, etc., the vast majority of studies, including the largest ones and the cleanest ones that really compared apples with apples (e.g. actual experiments or matching methods) found no gender bias in either direction.
Theynote that their overall finding contravenes the dominant narrative. which may explain how the paper was handled by the journal (see below):
Note how divergent these conclusions are from the dominant narrative that pervades the scientific media. Figure 2 appeared in Nature (Shen, 2013) and captures what many regard as the ground truth, namely that women in science earn 18% less than men and are far less likely to get funding.
The funding claim isn’t supported, and while there may be a bias-induced difference in salary, it’s more likely to be closer to 4% than 18%. That still needs examination, though, and then fixing if it’s due to bias. Note as well that this paper isn’t yet published.
One reason it may not yet be published is in fact that the findings of Ceci et al. are politically unpalatable: every inequity must, says the dominant narrative, be due to bias. This is not just sour grapes, as the authors argue. This excerpt, though long, is worth reading:
Our study was submitted for review at a top journal but declined by the editor, based on seven reviews, four of which recommended publication. It is interesting that, unlike our analyses of less controversial topics, whenever we have attempted to publish work on the underrepresentation of women in science that argued against a dominant role for bias, journal editors have felt the need to solicit many more reviews than is customary. We have seen this phenomenon often.
For example, in 2014 when two of us (WMW and SJC) submitted a manuscript on hiring bias to the Proceedings of the National Academy of Sciences, the editor solicited seven reviews, whereas the typical number of reviews for that journal was at that time only two. Many other articles that we have written individually or together have gotten this kind of extra scrutiny when we find no gender differences, and this can be compared to the relative [sic] lower scrutiny we get when we find gender differences. A similar situation may very well have been the case for this Stewart-Williams and Halsey paper. While we do not know how many reviews this article had, we know that it was submitted to several other journals because we or people we know were reviewers.
Perhaps an uncommonly large number of reviewers is appropriate when a manuscript challenges the dominant narrative that sex differences in academic outcomes is a consequence of gender bias rather than non-bias factors. Such a position goes against some reviewers’ “priors,” and therefore one could argue it is in need of stronger evidence than a claim that is congruent with the dominant narrative. Certainly, we would want a larger than usual number of reviews if a manuscript purported to provide evidence that validated ESP or voodoo, because such claims go against our deeply held beliefs that are based on decades of empirical and theoretical evidence.
However, what body of evidence leads to such deeply held beliefs that would require an alternative argument to findings such as ours showing no bias in tt hiring or Stewart-Williams and Halsey’s evidence of preference-based and perhaps biologically based career choices? What body of evidence would render such findings so aberrant as to require extraordinary evidentiary vetting? Note that we are not arguing that informed scholars cannot criticize these arguments. They indeed can, and should. Rather, we are arguing that in view of the scientific evidence they bring, why would Stewart-Williams and Halsey’s paper, or ours on lack of hiring bias, be so unbelievable? In light of the evidence on equal success rates for grant applications (both NIH R01s and NSFs in all of its directorates) for so many years, why do so many researchers continue to cite a 1997 article on gender bias at the Swedish Medical Council that—if ever there were gender differences—had disappeared by 2004 as demonstrated in a less cited but methodologically superior paper (Sandström & Hallsten, 2007)?
What would it take to get critics’ priors into sync with the published empirical data, when that data indicates no bias?. . .
By the way, I have no idea whether the Stewart-Williams and Halsey paper was given a harder review than normal given its conclusions; the authors say nothing about that.
Ceci et al.’s conclusion:
We believe that we can come to a deeper understanding of the causes of the differences in women’s representation in STEM if people drop their priors when evaluating evidence.
Dropping priors—that is, sitting down before the facts, as Huxley said, like little children—and remaining objective instead of trying to find data supporting your preconceptions—these are sine qua nons in scientific behavior. It is odd that scientists in this case are so clearly critical of data that go against their preconceptions and yet so willing to accept data that support them. We’re human of course, but we’re supposed to be fighting against our confirmation bias. That means giving all papers equal scrutiny, not extra scrutiny to papers whose results you don’t like. In fact, if anything, we should be giving more scrutiny to papers whose results we do like, or which support our biases.
________________
Stewart-Williams, S. and L. G. Halsey. Men, women and STEM: Why the differences and what should be done? Eur. J. Personality 35:3-39.
Ceci, S. J., S. Kahn, and W. M. Williams. 2021. Stewart-Williams and Halsey argue persuasively that gender bias is just one of many causes of women’s underrepresentation in science. Eur. J. Personality 35:40-44.
I’m on another Feynman bender, and in Six Easy Pieces, the editor notes that Caltech admitted only males as of about 1960. I found that striking – rather late to get with the program, Caltech.
Most of the Ivies didn’t go coed until the late ’60s (and Columbia took till ’83).
Louise Perry (a young British journalist and secular feminist) has just published a book entitled The Case Against the Sexual Revolution, arguing, inter alia, that men and women do differ, on average, in certain attitudes and behaviors. It will be very interesting to observe the reactions in the media to her claims.
“We believe that we can come to a deeper understanding of the causes of the differences in women’s representation in STEM if people drop their priors when evaluating evidence.”
I rather like that quote and I suggest one could insert just about anything political, social, or economical, much less scientifical, into the bolded section.
The narrative that all occupational disparities between the sexes or between racial groups are entirely due to “systemic” prejudice became dominant for reasons that have nothing to do with empirical tests. The narrative comports with an ancient superstition of the vulgar Left, namely that uniformity is the same thing as equality.
The narrative’s spread in the US has a sociological history. (1) It is embedded within “the culture of complaint” analyzed 30 years ago by Robert Hughes in an excellent book. (2) It provides a pretext for employment and make-work of a bureaucracy set up first in academia, and then in private corporations. (3) In the corporate world, the now dominant narrative provides a convenient way to distract attention from a different trend: the spectacular and continuing increase in the ratio of compensation at the top levels to pay lower down. One notices that there are no bureaucrats at all and few scholars who concern themselves with “systemic” feudalism.
“The narrative comports with an ancient superstition of the vulgar Left, namely that **uniformity is the same thing as equality.**” (my emphasis)
The emphasized part is especially well put.
Seconding Jon’s endorsement of Robert Hughes’s book, The Culture of Complaint. Hughes died almost exactly ten years ago. Oh, if only he and Hitch were still alive today…!
On “dropping priors,” that’s fine if it means biases and unproven assumptions. However, “priors” has become a technical term in the Bayesian analysis so prominent the last few years. Applying priors – everything you actually know – is what makes Bayesian analysis so powerful. For the topic under discussion, important priors include differential birth rates, expected lifespans, and good data on interests.
When I began reading this blog post my first thought that one of the major differences was having kids. Different demands are put on fathers and mothers, and those demands make it more difficult for mothers in some STEM fields. There’s no way around it. It isn’t impossible for women to have very productive careers in research heavy career paths, but it’s a lot harder for mothers than for fathers, at least from what I have seen.
Compare the number of childless men and childless women.
We may look forward to a directive that men should in future have 50% of all childbirths. The directive will be issued jointly by the Associate Dean for Title IX, and the DEI Committee for the Promotion of Virtue and Prevention of Vice.
This is one very long article(!), which I only now completed reading in its entirety. It’s interesting! Two things that stand out to me in reading it. First, it’s great to read an article that questions the thesis that sexism is solely or mostly responsible for the dearth of women in STEM professions. Let’s hope that this article is joined by others and that a real conversation can take place regarding this issue. This article may not be exhaustive—an exhaustive treatment of just the existing literature alone would require a whole book, the authors note—but it covers a lot of territory. The second thing that struck me is how careful the authors are to qualify every sentence. This tells me that they are fully aware they are taking a risk in publishing their thesis and that they are trying to preempt the purposeful misunderstandings that might be meted out by critics. Perhaps their careful approach will help them get a fair hearing.
Thank you for bringing this article to our attention.
I saw this coincidental tweet today, and it was not from James Damore !!
“Why are women so much more represented in astrology than in other areas of physics?”
One of the unspoken assumptions in all of these debates is that STEM jobs are highly desirable. I think this assumption is highly questionable.
Philip Greenspun:
This article explores this fourth possible explanation for the dearth of women in science: They found better jobs.
Why does anyone think science is a good job?
The average trajectory for a successful scientist is the following:
age 18-22: paying high tuition fees at an undergraduate college
age 22-30: graduate school, possibly with a bit of work, living on a stipend of $1800 per month
age 30-35: working as a post-doc for $30,000 to $35,000 per year
age 36-43: professor at a good, but not great, university for $65,000 per year
age 44: with (if lucky) young children at home, fired by the university (“denied tenure” is the more polite term for the folks that universities discard), begins searching for a job in a market where employers primarily wish to hire folks in their early 30s
This is how things are likely to go for the smartest kid you sat next to in college. He got into Stanford for graduate school. He got a postdoc at MIT. His experiment worked out and he was therefore fortunate to land a job at University of California, Irvine. But at the end of the day, his research wasn’t quite interesting or topical enough that the university wanted to commit to paying him a salary for the rest of his life. He is now 44 years old, with a family to feed, and looking for job with a “second rate has-been” label on his forehead.
https://philip.greenspun.com/careers//women-in-science
Don’t Become a Scientist!
Prof. Jonathan Katz, Wash. U. physics
http://yangxiao.cs.ua.edu/Don't%20Become%20a%20Scientist!.htm
^^^that is the correct url but it has to be copied/pasted.
Almost forgot – with the Nobels coming up, was reminded :
No discussion of academic science careers is complete without the story of Douglas Prasher – I think this part is relevant at the age 44 mark in Greenspun’s piece :
“Prasher had applied to the National Institutes of Health for funding but had been turned down, and by the time he was undergoing review for promotion from assistant to untenured associate, he had decided to leave academia.[citation needed] […] … unable to find a job in science, his life savings had run out, and he was working as a courtesy shuttle bus driver for a Toyota dealership in Huntsville at $8.50 an hour. [6][11][13][14][15]”
https://en.m.wikipedia.org/wiki/Douglas_Prasher
It’s good to examine all the possibilities of why women are less successful in STEM disciplines. When all the research is done and the woke blank slaters are driven back, I’ll bet my house it becomes clear that women are – on average and at the upper extremes – not as good at math as men, and that the difference is rooted in biology, not policy. And this explains the differential success in STEM as a whole.
This is very interesting, and thank you for publicizing it.
Ceci et al note that they did find gender bias in students’ ratings of teaching skills. Last year I commissioned an analysis of *all* student evaluations of teaching in the Faculty of Science at my NZ university. The results were that female teachers were rated slightly more *highly* than males in 8 of our 10 departments and schools on the overall teaching effectiveness question, and slightly lower in the other two. None of the differences in ratings was significant, and neither was the overall difference. There was also no discernible difference between departments with higher (e.g., Psychology) and lower (e.g., Computer Science) levels of female representation among staff and students. All this was true for two semesters’ data, analysed independently, providing some evidence of repeatability of the finding. One of the last things I did as Acting Dean of Science, before stepping down over the matauranga Maori controversy, was to present this analysis to University management.
Those of you who are in a position to do so might conduct a similar analysis at your own institutions. If there really is bias in students’ evaluations of their teachers, we need to investigate it further, and try to change it. But if there isn’t, we do our female colleagues no service by encouraging them to believe that the cards are stacked against them and that they’re doomed when it comes to, for example, promotion. Deciding on the right course of action requires solid information on the current state of play.
It may also be that a bias emerges under some circumstances but not others.
Perhaps, there are differences between societies that have different experiences of female leadership. In NZ, it’s common at the political level, and has been for many years. In the last national election, for example, the leaders of both major parties were women, and this attracted essentially no comment in public debate. We are no longer surprised by the existence of female authority figures.
It’s not straightforward to push back against bias-assuming narratives that have become entrenched in academic discourse, and providing data may not do the trick. I met the argument that ‘because we know there’s gender bias, it must actually be that the female teachers are much better than the males and it’s the bias pulling this back to apparent equality.’ Possibly, though Occam’s Razor suggests that it’s not the explanation we should endorse first.
Research like these two papers at least moves the issue to an empirical rather than ideological footing – thank you again for bringing them to my attention.
I think I’m missing something here. If ratings for men, as opposed to women, on the faculty differ (or are the same) how can that tell you anything about bias (or lack if it)? Suppose all the women on the faculty are in reality excellent and all the men mediocre, would your analysis find bias if the student ratings were objective (i.e. all the women got excellent evaluations and all the men mediocre)?