Wednesday, October 12, 2016

Is Sexual Harassment Rampant in Astronomy?

Jackie Speier has proposed legislation to tackle what she describes as "rampant" sexual harassment in the sciences. As an example of the sort of thing she is talking about she cites Geoff Marcy's treatment of Sarah Ballard. Before I continue my reflections on his behavior, I want to pause and put this data point into a larger perspective.

On Sunday, I described my encounter with the CSWA Survey of Workplace Climate. For the reasons I outlined in that post, I give very little credence to the study. It's not just the way the data was collected, but the way it was presented. It's not just the presentation of the results, but the authors' refusal to discuss those results. Today, I want to draw attention to three studies that are credible in an equal and opposite sense. Not only are all three studies written up and published, their authors have all responded thoughtfully and informatively to my inquiries.

In this post, I'm going to introduce the three studies briefly and state my main take-ways. These are not always the implications that the authors of the studies themselves emphasize, but they are, as far as I can tell, reasonable conclusions to draw. Like I say, I have conferred with the authors, but I take full responsibility for any errors I might make in what follows. If they are pointed out to me, and I understand the objection, I will of course correct the post (with proper acknowledgement). The comments section is a good place to have any substantive disagreements.

All three studies, as I read them, offer a more hopeful picture of astronomy (or in one case academia in general) than is currently being presented in the media. All three belie, as far as I can tell, the idea that sexual harassment is "rampant" in astronomy. I take this to be good news. It means that if you are an intelligent, curious woman who is interested in the stars and planets, you've got a relatively safe discipline in which you can seek employment. By "relatively" I mean mainly: compared to other places you might succeed if you're smart—finance, IT, and the military, for example. But I also mean that it's a safer place at 30 than it was at 20, and safer at 40 than it was 30. As Dan Savage wisely said: "It gets better."

Also, it's very, very unlikely that you will be sexually harassed in any very serious way in astronomy. Certainly, of the choices you make as an adult, getting into astronomy is not likely going to be a mistake if your aim was to avoid sexual harassment.

The first study presents fresh results from the The Longitudinal Study of Astronomy Graduate Students (LSAGS), sponsored by the American Institute of Physics. It is authored by Rachel Ivie, Susan White and Raymond Chu. In "Women’s and men’s career choices in astronomy and astrophysics," published in August, 2016, they argue that "the process of understanding attrition from astronomy and astrophysics must include multiple factors and cannot be reduced to a simple model in which respondents’ sex alone is the causal factor. The respondents’ sex had no direct effect on working outside the field." There were important indirect effects, it should be noted, but they don't seem big enough to warrant concern. For example, the respondents seemed generally very happy with their advisors, even if women rated them lower on average. The most striking result, to my mind, was that women don't seem more likely even to think about leaving astronomy than men. This makes the likelihood of finding a strong effect from gender-based harassment very low.

"We know that sexual assault and harassment are an enormous factor in driving women out of STEM," says Speier. But the truth is that there are no signs that women are being "driven out of STEM" at all, let alone that it is because there is an "enormous" amount of sexual harassment and sexual abuse in STEM fields.

At this point, let me pause to note a general point. In general, we would expect to find more harassment in a discipline than outright sexual assault, and we would expect to find more sexism than harassment. (Just as we would expect to find more casual sexism than blatant sexism.) My point with this study, and the next one, is that there would seem to be no notable "sexism effect" in astronomy. That is, there is very little of the thing that we'd expect to find most of in regard to these issues. If sexual harassment were "rampant" in astronomy then so, surely, would sexism. And if sexism were rampant in astronomy we would expect women to at least consider leaving the discipline more often than men. But this, Ivie, White and Chu found, is not the case.

The absence (or at least rarity) of sexism is also evident in a rather elegant study of time allocation at the European Southern Observatory.* In "Gender Systematics in Telescope Time Allocation at ESO", published earlier this month, Ferdinando Patat presents the results of an investigation of the relationship between gender and successful applications for telescope time. As one would expect, more men than women are granted telescope time, but this is of course largely because there are more male than female astronomers. There is also, however, a difference in success rates by gender: 16.0% for women and 22.2% for men. It seems men have a 1.39 times (39%) better chance of getting telescope time.

But one reason for this, which Patat was not able to fully correct for because the data couldn't easily be sorted for it, was the seniority of the principal investigator. As far as he could correct for this [using a very crude bin structure to represent seniority], he found that parity of success rates would give 19.3% for women and 22.2% [21.0] for men. That is, given what we do know about seniority and its effect of the success of an application, we would expect men (who are not just more numerous, but generally more senior) would succeed 1.15 [1.09] times more often than women (15% 9% more often) for reasons unrelated to their gender. That's much lower than the observed disparity of 39%. We don't know that this difference is accounted for by gender bias but, if all of it were, the upper bound to a "sexism effect" on telescope time allocation would be 24% [30%].***

The lower bound is effectively zero. That is, once the remaining effects of seniority [using a more granular approach to seniority] and other factors are taken into account, it may turn out that there is no sexism in time allocation whatsoever. That's not just a theoretical possibility, either. In the one group where seniority could be completely ruled out as a factor, namely, students, there was no gender disparity in success rates. That is certainly a hopeful result for young women going into astronomy.

Finally, I want to look at a study of actual harassment in academia. This one draws conclusions that are somewhat at odds with the ones I would draw, but the study itself seems well done and credible. It's not the numbers that I would question but their implications.

As the title suggest, "Still Second Class: Sexual Harassment of Graduate Students" by Marina Rosenthal, Alec Smidt and Jennifer Freyd, has an explicitly "feminist"** moral in tow. They found that female graduate students experience greater rates of harassment and perceive themselves as less safe than men. What stood out for me, however, is how mild the harassment seems to be in both groups. The mean score for faculty harassment of female students was 1.44 on a scale of 0 to 72. To put it in perspective, note that the scores are calculated using responses on a 0-5 [0-4] scale to various questions. If everyone had said they had experienced “sometimes” (score of 2) being “Repeatedly told sexual stories or jokes that [they found] offensive” and no other behaviors were experienced, then the corresponding mean score would have been 2. Given that behaviors are not, of course, spread that evenly, this does not seem like a population that is experiencing "rampant" harassment.

The study applies a method developed by Louise Fitzgerald and colleagues, originally to study harassment in the military in the 1990s. It is interesting to note that the comparable scores there were 10.45 for women and 2.39 for men. It may not be a perfect apples-to-apples comparison, but at first pass it would seem that women in academia today are less harassed than men in the military in the 1990s. It is interesting to think about whether that sets a high or low bar.

According to the study, female graduate students have a significantly lower perception of their own safety than their male counterparts: 3.36 vs. 4.32 on a scale of 0 to 5. Here we can ask whether the 4.32 is an acceptable number. I think it may well be, since we can't expect an environment to offer 100% of the people a complete sense of safety. Recall the harassment score: for men it is .59; for women it is 1.44. As the male score suggests, we probably have to decide on an acceptable but non-zero value. The same goes for safety, which must have an acceptable level that doesn't score 5. In both cases, unfortunately, we may also have to accept a slight disparity between gender within an acceptable range.

This raises a delicate issue: we are talking about perceived safety. How much, one wonders, could be gained by using the low harassment scores to promote a perception of safety? That is, is it possible that women in academia feel less safe than they really are? Might they perceive some behaviors as threatening though they pose no danger, whether physical or professional? This, indeed, is my major hypothesis in the Sarah Ballard case: she misinterpreted a friendship as harassment because she was primed to do so by an ideological framework.

A final note about "Still Second Class": the numbers seem to be driven somewhat by the experience of law students. That's not good for aspiring lawyers, to be sure; but it does suggest better conditions in STEM, and, of course, astronomy.

I will take these studies up again on their own terms in separate posts. I just wanted to bring them together first and emphasize my major take-aways.

*Update (15:26): I hadn't noticed that Science has already covered this result and spun it, predictably, in the other direction, describing it as further evidence of a problem and also linking it to the harassment issue. (HT Ferdinando Patat.) I'll take this up in a later post. To my mind, the important thing to notice is that the specifically gendered effect is probably quite small, and it may be the result, not of gender bias so much much as gender differences. There may, of course, be some actual sexism at work here. But, as I keep saying, I think the main message should be that women can safely go into astronomy and expect to find, on the whole, much less sexism in their way than they would in other vocations.

**These are scare quotes, not quotation marks. I just realized that my use of the word "explicitly" might make it seem like Rosenthal et al. actually use the word feminism. They don't. But the paper is published in Psychology of Women Quarterly, and seems quite open about its feminism to me.

***(Update 13/10/2016 at 9:45): Thomas Presskorn (in the comments) was rightly puzzled that correcting for seniority predicted the observed value exactly. I had actually gotten this wrong in two ways. Patat's paper says 22.1 (not 22.2) for the predicted value. In my email exchange with Patat I had misunderstood how this correction works and assumed that the observed male success rate served as a baseline. Discussing it further, however, Patat realized that the 22.1 figure is actually a typo, and should be 21.0. This makes the situation a bit worse than I had suggested, moving the upper bound of the sexism effect from 24% to 30%. I will discuss all this at greater length on Friday.


Presskorn said...

Welcome back to blogging. A few non-expert yet technical comments:

1. Daniel Dennett once said that when people use the words "surely" and "obviously" something is, more often than not, fishy about their inferences. This leads to ask inquire about the justification for "If sexual harassment were "rampant" in astronomy then so, surely, would sexism." - Why exactly would that "surely" be the case? ...

2. "As far as he could correct for this, he found that parity of success rates would give 19.3% for women and 22.2% for men." - Shouldn't the number for men also change if corrected for seniority? It seems weird that it continues to be 22,2%? Or perhaps ALL the male applicants were senior?

3. "According to the study, female graduate students have a significantly lower perception of their own safety than their male counterparts: 3.36 vs. 4.32 on a scale of 0 to 5. Here we can ask whether the 4.32 is an acceptable number."... Is there a typo here? Should the last sentence have asked whether the female score of 3.36 is an acceptable number?

Thomas said...

Hi Thomas, nice to have you along!

1. It's my understanding that all sexual harassment is considered gender-discrimination. On that view the amount of sexism in an environment will always be greater than or equal to the amount of harassment. So we can be sure that if harassment is rampant sexism is too.

2. I think the male success rate sets the baseline. The 19.3 is the downward adjusted expected outcome (success rate) given what is known about the composition of the female population. (I'll check this with Nando before writing the next post on this study.)

3. I was trying to say that 4.32 is probably an acceptable score for perceived safety. The important thing is that it is not 5. That raises the question: what non-5 score for women would be acceptable? How low would we allow the male score to get before being alarmed by it, too? Surely (!) it would be an implausible stroke of luck that men feel just exactly safe enough for our approval?

One of the weaknesses of the study, and indeed the policy-environment around sexual harassment in astronomy, is the absence of threshold values. Even one case of (any kind of) harassment, it is said, is one too many. There is a sense in which that's true, of course. But it doesn't tell us anything about what a reasonable policy response would be.

The same problem can be replayed, actually, on the question of disparity. Might it be acceptable that women, as a group, feel slightly less safe than men? Might that be a fact of nature we can't completely eradicate from any profession? I leave that as a question.

Anonymous said...

With regard to point 3, I think it's fairly well established in the social science literature that women generally perceive themselves as being less safe than men. This is despite men being the overwhelming majority of victims for crimes like robbery, assault, and murder. Of course some people say that because women perceive themselves to be in more danger, they take more action to avoid becoming a victim. It certainly doesn't hurt, but is that really true, or is it just an overreaction to a threat that doesn't really exist?

Thomas said...

I haven't looked at the literature on that specifically. But you raise a good point. It would be nice if these surveys began with some sort of hypothesis. On a scale of 0 to 5 how safe would we expect men feel on average in astronomy, in finance, in teaching, in nursing? How safe would we expect women to feel in those fields? Would we expect (or demand) that women feel as safe as men in all fields?

I've been framing those question in commonsense terms. But it's a good idea to think of it in terms of the background knowledge that social science provides. Thanks for that. I'll see if I can find an answer.

Anonymous said...

You raise a valid point about the fact that it is not known who should feel more safe as a group, men or women. Men may be more likely to be victims of robbery, assault and murder, but women are more likely to be victims of rape, sexual assault and harassment. And in the work environment - which is a more likely class of crime ? So women being more afraid in the work place is not entirely unreasonable.

Your overall point that I take away, is that we have to have some acceptable base level. 0 is totally not possible, and there is economic theory that shows why this is the case. So to discuss a problem, we need a ground state to compare it to. That may be different for men and women and naively that seems reasonable.

Presskorn said...

A small comment on your response to 1. Yes, IF we define sexual harrassment as a proper subset of gender-discrimination, then we can indeed be SURE.

But really, we shouldn't do that. Treating sexual harrassment as a proper subset of gender-discrimination will, in any case, make us more liable to take evidence for gender-discrimination as evidence for that the fact that, e.g., Marcy did really sexually harass Ballard. And vice versa, if Marcy did really sexually harass Ballard we will be prone to see that as evidence of "rampant sexism" in STEM. Both evidential inferences, however, are obviously false.

Thomas said...

I'm inclined to agree with you on the principle. Sexual harassment is possible even without the harasser being a sexist. (And being a sexist doesn't make you a harasser.) The motives and powers involved are often quite relational and situational, i.e., specific to persons.

(It's interesting that when you read Speier's bill, it doesn't say anything about harassment. It applies to any discrimination.)

But I'm also inclined to agree with those who see an empirical correlation between sexism and harassment and I sympathize to some extent with those feminists who, in the early days, saw harassment as a "gender issue". I think it is often true that a harasser feels free to behave as he (!) does because he has a particular view about the role or place of women in social life.

Interestingly, I think something just as disturbing is going on on the other side of this issue. The destruction of the lives of casual sexists -- and even non-sexists mistakenly taken to be sexists -- under the banner of fighting "harassment" and even "assault" depends on a certain view of men, and, I think, an overly strict enforcement of "professionalism" in the workplace. Some have called this a "feminized" enivironment, but I think this places the blame in the wrong place (namely, on women). It's actually a corporatized environment. Feminism has just been a useful ideology to move this process forward.