Monday, March 17, 2014

Do Women Have an Advantage in Faculty Searches?


Ahh, we are halfway through the month of March. There is more sunlight. The Kepler field is rising. And, the faculty search season begins to bear fruit, and we learn (often through the Job Rumor Mill) who has received a faculty offer.

The faculty search process is enormously stressful to the postdoctoral fellow community, who are competing in a very oversubscribed market. Every March I hear this frustration boil over (expressed perhaps as an off-hand remark over coffee, or in a snippet of conversation I catch in the hallways) with that most poisonous of analyses: Dr. So-and-so received the job offer because she is a woman.

It would be wrong for me to imply that I hear this only from postdocs: Certainly in my travels I have heard senior scientists and graduate students imply this as well.

The fact that anyone remotely familiar with the statistics could assert that it is somehow a net advantage, from an advancement perspective, to be a woman in astrophysics seems ludicrous to me, yet I do hear advocates for this idea without fail each year. So, let's take a moment to remind ourselves of the data.

Recently, Prof. A. Meredith Hughes (Wesleyan University) led the Committee of the Status of Women in Astronomy on its third demographics survey. The CSWA had conducted previous surveys in 2003 and 1999, and a similar survey had been first conducted by astronomers at the Space Telescope Science Institute in 1992. So, that means the baseline is just over 20 years long, and importantly the participation rate is nearly 100%. The CSWA survey is unique in that it is an attempt to survey all professional astronomers in the US, both those in academic departments and those at research institutes. (Note the AAS has also just published its very informative survey, but the CSWA survey is distinctive since it does not require AAS membership.) The CSWA survey data are fascinating, so if you haven't yet read the previous post about the results, I encourage you to take a moment right now. Here I wanted to look at what the survey data have to say about the question of whether being a women is, in itself, a net advantage in moving from the graduate student level to an assistant professorship.

The analysis of the survey data compared the fraction of assistant professorships held by women with  the fraction of the graduate student pool who are women, separated by ten years. (The duration of each career stage is roughly comparable at 5-7 years, and the elapsed time to move from graduate school to a junior faculty spot is also roughly 10 years.)

OK, so first let's look at the 1992 graduate student and 2003 assistant professorship cohorts. Between 1992 and 2003, 30 +/- 3 % of graduate students who were men advanced to the junior faculty level, whereas 18 +/- 3 % of graduate students who were women advanced similarly. Hmm, I do see an advantage there, but not for the women! Keep in mind that for at least 20 years the fraction of prize fellowships awarded to women has held roughly constant at 30% (Schmelz et al. 2010). Now, let's turn to the most recent decade: Doing so, we find that the large difference between men and women has vanished, with advancement rates of 19 +/- 2 % for men and 16 +/-3 % for women (statistically indistinguishable) between the 2003 graduate students and the 2013 assistant professors.

So, the summary of the data is that:
  1. In the past, your chances of advancing from graduate school to a junior faculty job were much worse if you were a women.
  2. In the past decade, this disadvantage has disappeared, but there is no evidence that women have a net advantage or experience a higher rate of retention or advancement.
(Of course the third conclusion is that the overall rate of advancement has declined for both men and women. As Prof. Hughes wrote in her summary, the job market is bad for everyone.)

Now, I have served on many search committees, both inside and outside of my department, and in most cases there is, at some point, a direct discussion about the desire in increase the participation by women. And largely (but certainly not uniformly!) I have been impressed by the thoughtfulness of my colleagues on this issue. Why do I mention this? Because perhaps it is during those discussions that this perception of tilting the playing field to the advantage of women finds its footing. Or perhaps it also occurs when an former advisor sees his or her male graduate student lose out in a faculty competition. Or perhaps it occurs when applicants themselves see short lists with a gender breakdown very different than that of the senior faculty itself. Regardless, the data are very clear: For the past decade, faculty searches have largely hired women and men with relative rates that match their availability in the graduate student pool a decade earlier. This didn't used to be the case, but that mistake has now been corrected.

I am writing to ask that we all take care to avoid unfounded speculations about the reasons why a particular individual was selected for a particular job offer, and that we be meticulous about not attributing the hard-won success of applicants to their gender. I also invite you not just to avoid such speculations, but to correct such statements when they are presented to you. Claims that an individual won a competition based not on his or her achievements but rather on his or her innate characteristics are poison to our community. They are particularly toxic to the new faculty members themselves, who would rather not like to begin their jobs having to disprove such unfounded rumors.

9 comments :

Anonymous said...

Interesting article, but the data presented here do not directly address the question posed ("Do women have an advantage in faculty searches?") To address that question, one would have to compare the fraction of faculty job applicants that are female with the fraction of successful applicants that are female.

As this blog has noted previously, the rate of attrition is greater for females at each step of the 'pipeline'. Consequently, it is likely that the percentage of graduate students that apply for faculty jobs (after one or more postdocs) is less for females than for males. If so, then the data presented in this article demonstrate that of those (former) graduate students who chose to apply for faculty jobs, females are more likely to receive a faculty job.

This effect is illustrated in the first infographic of Shen 2013, Nature, 495, 22–24 (http://www.nature.com/news/inequality-quantified-mind-the-gender-gap-1.12550). (Note that these data are for physics, chemistry, and biology, not astronomy. It would be interesting to see similar data that are specific to astronomy.) The caption states the conclusion quite clearly: "Female representation among science and engineering faculty members in the United States has lagged behind gains in graduate education, in part because women do not apply for tenure-track jobs. But women who do apply are more likely than men to receive interviews and offers."

We can speculate about the reasons for the greater rate of success for female candidates, but the simplest explanation (that I can think of, at least) is that hiring committees act intentionally to combat other factors that lead to under-representation of females amongst junior faculty.

None of the above justifies the statement that "Dr. So-and-so received the job offer because she is a woman", nor does it address whether such 'positive discrimination' should be employed by hiring committees. It simply demonstrates that there is evidence that the question posed by this blog post should be answered in the affirmative, and it is disingenuous to argue otherwise by presenting data that do not directly address the question posed.

Seth Zenz said...

Dear Anonymous,

This is an interesting additional dataset and an interesting piece of information, although it goes too far to say it's "disingenuous" to look at faculty hiring as part of long-term evolution rather than just the hiring process itself -- it is, rather, one part of the puzzle, and the part that the data being looked at were able to address. But in any case, the article you link doesn't at all support your conclusion that "the simplest explanation . . . is that hiring committees act intentionally to combat other factors that lead to under-representation of females amongst junior faculty." A perfectly reasonable (equally "simple"?) alternative speculation is that, with so many women discouraged from applying for faculty jobs in many ways over many years, the ones who remain are better qualified on average. The truth is, it's hard to actually answer the "unfair advantage" question itself with data. We would need to have the "real" answer about who was "best qualified" and "should" have been hired in each faculty search, and that doesn't really even mean anything!

Unknown said...

Dear Anonymous,

Thanks for the comment. I agree it would be interesting to gather data on the make up of the applicant pool. I think it would also be good to gather data on offers, which is not the same as who ultimately took the job. But I disagree with your assumption that the correct pool to examine is the pool of applicants.

There are two reasons:

First, bias in hiring might not be occurring solely at the point of selecting the person to receive the offer, but may begin much earlier (membership of the search committee, who is encouraged to apply, etc). Related to this, perhaps one group is better at identifying whether they are viable candidates than another. Regardless of the reasons, it could be that the average competency of the two groups in the applicant pool isn't the same, whether due to external pressures or self-selection. But I don't think it would be correct to conclude that one group has an advantage just because their success rate exceeds their fractional make-up of the applicant pool. Said differently, you speculate that hiring committees are acting intentionally to favor women, but perhaps it is merely that the women who actually apply are, on average, stronger than the men who apply, for the reasons listed above. The second reason is simply that some schools do not conduct open searches but approach a short list, or even a single individual, directly from the community. Obviously in those cases there is no list of applicants to begin with.

The reason for focussing on the graduate student to faculty attrition rates is not, as you accuse, to be disingenuous, but rather to identify a stage through which all prospective faculty must pass: A professorship requires a PhD. The intermediate product, the list of applicants, is not always available and may already reflect many biases.

Anonymous said...

Statistics are great, but do not help in understanding the hiring process on a case-by-case basis. Globally, it shows the trends, but I think we can all admit that discrimination occurs both ways. There are hires where females are discriminated against and that is absolutely reprehensible. However, there are admissions where the final offers are not just influenced by the qualification of the candidate, but also to increase diversification of a department. It could well be that so-and-so was hired not just because she was qualified, but also because she was a female. It could also happen, that there was a better qualified male for the job. If that is the case, it is equally reprehensible for him to be discriminated against. Can the contributors to this website admit to that?

Anonymous said...

Dear Seth and David,

Thank you for your thoughtful comments.

One thing that is clear to me is that we have defined the specific question/measure of fairness differently, which is important (and perhaps unsurprising). Based on David's introduction, I was considering 'fairness' from the point of view of a postdoc applying for faculty jobs and solely considering statistics (not speculation about the relative qualifications of males and females). In contrast, others, including David, have considered 'fairness' as the relative likelihood of progressing from PhD completion to a faculty job.

Thus, we clearly interpreted the question/fairness differently. Furthermore, David explicitly disagreed that the pool of applicants is the correct pool to examine. I find the stated reasons interesting and plausible but unconvincing (as presented, at least); I address each below.

Reason 1: The reason that I wrote "the simplest explanation (that I can think of, at least) is that hiring committees act intentionally to combat other factors that lead to under-representation of females amongst junior faculty" is that it's a plausible mechanism to explain the higher success rate of female applicants for which I am aware of some anecdotal evidence (first-hand, second-hand, and through posts such as John Johnson's recent posts about 'diversity hires'). However, because of the lack of actual data (I, too, do not base conclusions on anecdotal evidence), I tried to make it clear that this was only a speculation that one could make from the data presented (and not a conclusion). The primary point of my post was to present data that could be considered to contradict David's conclusion.

That female applicants are typically more qualified than the males is another plausible mechanism that has been suggested by Seth, David, and others. However, especially considering the message of the original post, it should be recognized that this qualification difference is a speculation for which no supporting data have been presented. In principle, some (certainly imperfect) quantitative measures of qualifications could be analyzed: do female applicants have more citations? Higher h-indexes? More papers? etc. Such measures are imperfect because of, e.g., implicit bias. Nevertheless, an argument based on such data (perhaps crudely corrected based on the results of studies about implicit bias?) would be more useful than simply asserting that gender-based differences in qualifications explain the higher success rate of female applicants.

Furthermore, if I wrote that "perhaps it is merely that the [men] who actually apply are, on average, stronger than the [women] who apply", I imagine that it would not be well received by many readers, and for good reason. For the sake of argument, I could list possible mechanisms that could result in male applicants being, on average, more qualified than female applicants, but I would not do so without strong supporting data.

My general point is that we can speculate about reasons for the higher rate of success of female applicants, but we cannot conclude anything without data such as I suggest above and, ideally, information about how faculty searches are performed from those who perform them. Unfortunately, the closed-door and confidential nature of these searches likely prevents us from moving beyond anecdotes regarding search committees explicitly seeking or heavily favoring female applicants.

Anonymous said...

Reason 2: The importance of this effect depends on what fraction of faculty searches are closed (and how many of them are closed and specifically target females). As I am aware of no data regarding such searches (and I imagine that it would be significantly harder to procure than data about normal searches!), I can't judge how important this effect is. Even if many searches are closed, data regarding applicants for the subset of searches that are open would still be useful.

Finally, regarding David's comment, "The reason for focussing on the graduate student to faculty attrition rates is not, as you accuse, to be disingenuous, but rather to identify a stage through which all prospective faculty must pass. A professorship requires a PhD." This is also true of a BS, high school diploma, etc. Shall we judge the "fairness of faculty searches" by considering the relative fraction of high-school graduates that become junior faculty in astronomy? I've argued that if we wish to consider the 'fairness' of faculty searches specifically, not the overall pipeline, we should be concerned with the pool of applicants. If that is undesirable because of the reasons mentioned by David and others on the Facebook group, then we should at least consider the pool of qualified applicants. How astronomy junior faculty are hired directly from graduate school? Almost none. Thus, those who complete at least a single postdoc would, in my opinion, better represent those with the highest required qualification for a faculty job.

Thanks again for the thought-provoking post and discussion.

Anonymous said...

Two comments, in response to anonymous comments above:

1) If I may put a few words in Dave and/or Seth's mouth, I think one implicit assumption in their arguments is that women are, on average, equally talented and equally interested in scientific careers (in a biological sense, cultural factors notwithstanding). Therefore when we see a field, or a level within a field, in which women are underrepresented, when we ask "why?" what we are really asking is "how can we prevent the disproportionate loss of women, which necessarily represents a loss of talent, since men and women are equally talented?" That's actually the reason why we would not agree with a statement like that suggested by anonymous above, to the effect that male candidates are overrepresented at the faculty level because they are more talented on average. The underlying assumption is that if women are disproportionately leaking from the pipeline at any stage, there is a loss of talent. Assuming that the cause for the leak does not actively select the most talented women (which none of the proposed mechanisms that I'm aware of do), then the women who persevere will be on average more talented and persistent than the ones who leaked. As Dave and Seth point out, there are many reasons to expect that the applicant pool itself is a biased sample, and the biases that prevent women from applying in the first place are relevant.

2) I think the most recent anonymous commenter goes a little too far (why look at the pool of graduate students? Why not go all the way back to the high school level?). To that I would argue that once you've received a PhD in astronomy (which is related to the supply of graduate students), you are probably sufficiently talented and motivated to become faculty (whereas a high school or undergraduate degree is necessary but not sufficient). Obviously not all graduate students are, but if we assume that female PhDs are on average as talented as male PhDs, then we would hope that the assistant professor pool would look like the graduate pool from 10 years previous (which is what the data are actually examining). Sure, there are lots of things that might happen in between... but the point is that if those things are causing women with PhDs to disproportionately leave the field, then we are losing talent, and that's a problem.

Anonymous said...

This is a fabulous discussion, and I can’t tell you how glad I am to hear it, because I often have these discussions one-on-one with people, but it is much nicer to hear many different viewpoints. As a recently graduated female on the job market, I would like to start my comment by saying I am biased ;). And I am biased because the absolute last thing I would like to hear when I start a new job is for someone to say that I was only hired because I was female. I would be ashamed, frustrated, mad, intimidated, and no one wants to spend their emotional energy or time on unwarranted rumors rather than doing their work.

It is hard for me to read people’s comments, arguing that for whatever reasons, however valid, that they don’t believe statistic XYZ that women are not artificially favored in job searches, because to me that sounds like it is ok for people to say that so-and-so was hired for reason X, where X does not equal merit, until 100 studies prove that person was not. Even if it is not about gender, no one wants to hear people say they were hired for some reason other than merit. Maybe there are some situations in which this is true, but as a community I believe we strive to be merit based. And even if we don’t in practice, that is the gold standard that people judge other people’s decisions on.

I would like to come back to the author’s point that saying such things is toxic. Lets assume for the moment, that this comment is untrue (the person was not hired for reason X, where X is not merit). I can’t really think of how anyone would benefit from someone making this untrue statement, but I can think of lots of ways that it can hurt multiple people and the department. The people who didn’t get the job don’t benefit, the person who got the job doesn’t benefit, anyone else with characteristic X doesn’t benefit, and if the person hired is meritorious and leaves the department then the department looses out too (not to mention all the time, energy and money that will be spent to hire someone else, or perhaps the department will just loose the position).

If the comment was true, the only possible benefit I could see is if for some reason the search committee “saw the error of their ways” and next time hired someone who was meritorious over someone who had characteristic X and that lead to a more productive contribution to the department. If the comment is true, then there are still the harms for the individual, others like them, and possibly the department if that person could have been more productive if they weren’t spending time dealing with the comment. In this competitive market, it seems relatively unlikely that anyone who is being hired has no potential or merit.

But again, this all depends on being able to objectively say that one candidate is better than another, which is sometimes apparent if there is one candidate who stands out way above the rest, but is usually pretty hard to do.

Anonymous said...

Dear Anonymous @March 21, 2014 at 2:04 PM,

Regarding your first point, "the women who persevere will be on average more talented and persistent than the ones who leaked" is true only if the less-talented women are are disproportionately leaking, right? My point in my post from March 18, 2014 at 8:10 PM is that one can think of mechanisms (e.g., family-related ones) that would be talent-neutral. I've even heard the (perhaps tongue-in-cheek) suggestion (from a female) that more-talented females leave disproportionately because they rationally assess the costs and benefits of an academic career and decide against it. Thus, unless data are presented that suggest otherwise, it is an assumption to state that 'more-talented' females are more likely to continue along the academic pipeline. (I do agree that those who continue are likely more 'persistent', perhaps by definition, but this is also true for males.)

Also, there are possible reasons to question the implicit assumption that you identify (see http://www.psychologytoday.com/blog/the-nature-nurture-nietzsche-blog/201402/the-sticking-point-why-men-still-outnumber-women-in-sc for a rather thorough discussion), but I don't want to get into that extremely emotionally and politically charged can of worms.

Regarding point 2, I am aware of numerous examples of individuals who performed well in graduate school - perhaps because of a very hands-on advisor, good projects being given to them, etc. - but were not successful at doing independent research as postdocs. Thus, I would argue that receiving a PhD is not sufficient evidence that one is qualified to be a faculty member, and the manner in which almost all faculty are hired (after they have 'proven' themselves through one or more postdocs) supports this claim.


Dear Anonymous @March 24, 2014 at 10:24 AM,

I agree completely that it is never okay to say "that so-and-so was hired for reason X" because, as you state, the potential negative consequences outweigh the positive. As I wrote in my original post, "None of the above justifies the statement that 'Dr. So-and-so received the job offer because she is a woman'". Even if there is statistical evidence for some bias, conscious or unconscious, one can never demonstrate that the bias was the reason that so-and-so was hired (unless one personally witnessed the hiring committee state, "Well, we'd prefer to hire Dr. X, but we have to hire a woman, so we're making an offer to Dr. Y instead" or the search specifically targeted, e.g., females).