Post edited (9/28/2012), see below!
A new study was released, discussed by Ilana Yurkiewicz at her Scientific American blog, that seems to conclusively prove that gender bias in the sciences does exist. The article describes the challenges involved in studying this and how these challenges were overcome:
It’s tough to prove gender bias.
In a real-world setting, typically the most we can do is identify differences in outcome. A man is selected for hire over a woman; fewer women reach tenure track positions; there’s a gender gap in publications. Bias may be suspected in some cases, but the difficulty in using outcomes to prove it is that the differences could be due to many potential factors. We can speculate: perhaps women are less interested in the field. Perhaps women make lifestyle choices that lead them away from leadership positions. In a real-world setting, when any number of variables can contribute to an outcome, it’s essentially impossible to tease them apart and pinpoint what is causative.
The only way to do that would be by a randomized controlled experiment. This means creating a situation where all variables other than the one of interest are held equal, so that differences in outcome can indeed be attributed to the one factor that differs. If it’s gender bias we are interested in, that would mean comparing reactions toward two identical human beings – identical in intelligence, competence, lifestyle, goals, etc. – with the one difference between them that one is a man and one is a woman. Not exactly a situation that exists in the real world.
But in a groundbreaking study published in PNAS last week by Corinne Moss-Racusin and colleagues, that is exactly what was done. On Wednesday, Sean Carroll blogged about and brought to light the research from Yale that had scientists presented with application materials from a student applying for a lab manager position and who intended to go on to graduate school. Half the scientists were given the application with a male name attached, and half were given the exact same application with a female name attached. Results found that the “female” applicants were rated significantly lower than the “males” in competence, hireability, and whether the scientist would be willing to mentor the student.
This does seem to be a fairly solid indictment of academia for its sexism. Surely feminists everywhere will trumpet this study as a victory. After all, it proves that discrimination exists from the gate, and surely that’s responsible for every difference in the aggregate outcomes of males and females, which includes the fact that women overall earn less than men, are promoted less, and make up smaller fractions of certain fields, particularly in the “hard” sciences and engineering.
Well not so fast. First of all, few people would claim that discrimination doesn’t exist at all in the workplace and in academia. Indeed, I’ve always suspected that it indeed does exist, and does serve to help to hold women back. But, in the minds of people incapable of contemplating nuance (which is most people, unfortunately), the fact that discrimination exists means that it must be the only reason for these differences in outcomes, and any one who disagrees with this must be a sexist pig trying to keep women “in their place”. Just ask this guy, a man who was Watsoned before the man who is the term’s namesake.
With respect to the “discrimination only model”, here’s something to consider: Women aren’t the sole gender who are victimized by discrimination. Just ask men who want to work in early childhood education (also here).
This study’s design, while quite laudable, was actually suggested by Steven Pinker in The Blank Slate way back in 2002 (p. 354):
there is no doubt that women faced widespread discrimination in the past and continue to face it in some sectors today. This cannot be proven by showing that men earn more than women or that sex ratios departs from fifty-fifty, but it can be proven in other ways. Experimenters can send out fake resumes or grant proposals that are identical in all ways except the sex of the applicant and see whether they are treated differently.
It seems that these researchers have done just that. They found that men and women were indeed treated differently, and in fact, they found that this difference was rather significant:
So case closed, right? We know that discrimination exists, and it’s apparently rather egregious, so we should try to fix that, and raise awareness among academic departments about their latent and apparently unconscious sexism, right?
Well I’m afraid it’s not quite so simple. Now that we know this, the next question that comes to mind is why is this so? The study found that the sex of the individual making the decision did not matter. Women were equally discriminatory of other women as the men were. Why is this so? Most feminists will have an answer, probably involving some story about the legacy of the evil Patriarchy. But I’m not so sure that’s the case.
For one, a couple of things to consider: As Larry Summers was “Watsoned” for mentioning, men have a higher standard deviation in IQ, so there are proportionally more men at the extremes. At the high end, where prospective candidates for high-level careers and academic positions will be found, this means that men outnumber women, and this becomes quite a large margin at very high levels of IQ (Richard Lynn and J. Philippe Rushton have also claimed to have found that in adults, men seem to have a higher mean IQ than do women, but there is some evidence that this finding may be in fact due to attrition of males on the low end). Hence, a significant share of the difference in male and female outcomes, particularly at the high end, is due to the fact that there are more smart men than there are smart women.
As well, there are other issues, such gender-specific cognitive strengths (men having an edge in visuospatial ability) and differing interests (not the least being that autism-spectrum traits are more common in men, and lend themselves to interest in physical sciences and engineering). However, let’s ignore those for the moment and return to the IQ differences. The resumes in the study were equalized in all ways except for the gender of the applicants. The “subjects'” respective accomplishments were the same. Hence, it should be inferred that each male should be equally intelligent as their female counterpart. However, the males were preferred regardless. As well, the reviewers found concrete-sounding reasons for rejecting the female candidates. In other words, they seemed to perceive the women as being less competent, even though they objectively weren’t.
Or were they? I’m about to touch an “advanced” concept that many readers of my blog probably aren’t quite ready for. I will state that if you haven’t quite gotten comfortable with the notion that “pointing out the existence of group-wide differences in the averages doesn’t tell necessarily you about a particular member of a group, hence people need to be judged as individuals”, then you’re not going to be ready for what I’m about to discuss. The reality of the situation is that there are limitations to that advice. For one, while group averages don’t determine anything concrete about group members, they do affect probabilities. It is not statistically unreasonable to prefer a female nanny to a male one because the female is less likely to sexually abuse your children. Just the same, it isn’t statistically unreasonably to avoid a strange male on a dark street but show no such apprehension to a strange female: the male is more likely to mug you.
I shouldn’t have to explain why this is tricky advice for me, as a person of color, to give.
OK, but that’s discrimination against more or less unknown group members. However, in this experiment, these people were vetted via their qualifications, so that should serve to remove many of the unknowns. There is still one other issue, and that is regression to the mean. As La Griffe du Lion explains, any sort of test is likely to overpredict the performance of high-scoring individuals, and much more so for those from low-scoring groups. This is because high scores are more likely to be flukes than scores closer to the mean. And individuals from low-scoring groups, having a lower mean, are more to generate suspect high scores (the individual from the lower scoring group is more likely to have been having a good test day), as Henry Harpending explains. Granted, much of this effect should be neutralized by the fact that the hiring people had more than a single test score in this case (more data brings more accurate measurements). However, that doesn’t change the fact that, overall, these individuals’ experience with female students/employees has shown them to be, overall, weaker performers, and it this, quite unfortunately, colors their decision.
The second factor is an inescapable one and one that needs to be addressed if the solutions I’ve previously discussed to societal problems are to ever come to pass. That is that women have children. Female candidates are simply more likely to take time off and interrupt their studies/work to have children. And indeed, evidence shows that this greatly negatively impacts women’s career advancement (also here). This is a major reason that educated women put off having children, contributing to dysgenic breeding among women. Employers and universities do take this reality to account, and as we see, prefer male candidates when all else is equal.
As commenter redzengnoist noted, European countries like his native Denmark have reduced this friction by providing working mothers with strong legal protection, entitling them to maternity leave and pay. Hence, having children is less of a career-killer for working moms, and the fertility rate among the educated is higher. I know that this is an anathema to the anti-socialist right-wing U.S., but only by embracing such policies (as Steven Pinker also suggested in The Blank Slate) can such discrimination—and the problems of dysgenic fertility—be addressed.
There is a third factor that likely plays a role in gender discrimination. That is group cohesion. Different workplaces and academic environments tend to select for certain types of people, and hence the individuals in a particular workplace or academic department tend to be similar to one another. In certain fields, where gender ratios overwhelmingly favor one sex or the other, it can be hard for those who don’t fit the mold to fit in. It is this that is partly responsible for the perceived hostility that male early childhood teachers experience: they are simply not part of the group. The feeling of unwelcomness such outliers experience in such settings may not be (or may be) overt discrimination, but rather a lack conformity to the group norms. For example, engineers, in addition to usually being male, are also often conservative—indeed libertarian—and this may make it harder for those who don’t hold these political views to join these departments. This fact is simply an aspect of human nature and may be difficult or impossible to ever completely eliminate.
Edit, 9/28/2012: There is a fourth factor that possibly contributes the observed behavior. See this comment to the original post on Yurkiewicz’s blog:
There’s another possible explanation, in addition to the four listed, that could explain why everyone (both men and women) rated the qualifications lower when they were attached to a female name:
(5) The applicant has been the beneficiary of affirmative action, and therefore she did not have to jump through hoops of the same difficulty all along the way.
This is an absolutely rational way to interpret data when one knows that affirmative action exists. It certainly exists pervasively with respect to race. It’s less clear with regard to gender. In fact because more girls apply to college than boys, it’s reported that colleges seeking a 50% male class admit boys a little easier than girls.
But in science and engineering, there’s always female “under-representation” and always someone saying we’ve got to do something about it. There is always a push for affirmative action, to get more females! This has been true during my whole life, ever since I started in college and could observe it.
I remember my freshman year in engineering school. Usually the class was 20% female, but that year they tried increasing it to 25% female. They admitted an extra 5% of the class as females with somewhat lower standards, to get the percentage up.
By the end of the second year, just about all of that 5% of extra females had flunked out. And this was a very meritocratic place, if you were clever they didn’t care if you were a small green man from Mars.
Steven Pinker also made this point in The Blank Slate (p. 358):
Many women scientists are opposed to hard gender preferences in science, such as designated faculty positions for women, or the policy (advocated by one activist) in which federal research grants would be awarded in exact proportion to the number of men and women who apply for them. The problem with these well-meaning policies is that they can plant seeds of doubt in people’s minds about the excellence of beneficiaries. As the as astronomer Lynn Hillenbrand said, “If you’re given an opportunity for the reason of being female, it doesn’t do anyone any favors: it makes people question why you’re there.”
This is why I am against gender and race based preferences in employment and education. They don’t do as must justice to the beneficiary group as we think, and it serves to sow resentment among passed-over individuals as well as contribute to the impression that candidates from the beneficiary group are inferior.
Edit: See also A Success Story?