Monday, September 24, 2012

What's In A Name? Science Faculties Prefer John To Jennifer.

Kathleen Geier discusses a recent study about possibly unacknowledged gender bias among science faculty in the United States.  The study is important because it comes out at a time when the general discussion argues that all the bias against women in the mathematical and science fields is inside their own heads or based on their own choices to focus on children and housekeeping duties and so on.

The basic idea behind the study (pdf) is not new but it's still pretty neat:  Make up information about a potential job applicant, make sure that the information looks realistic to insiders in the field, and then send it out to several evaluators, half of which get it with a female first name for the applicant, half with a male first name.  The important point is that ALL the other information in the package is identical.

Thus, what the study really tests is the hidden information or stereotypes the female and male first names elicit in the judges. 

Why is this neat, in my opinion?  Because these studies control for all the possible real differences between potential female and male applicants.  Any average differences that the study finds must therefore be attributable to the gender views of the judges, whether acknowledged or unacknowledged.

This particular study, called "Science Faculty's Subtle Gender Biases Favor Male Students", by an interdisciplinary team of researchers at Yale University,  had this basic setup:

Here’s the study’s methodology: a group of researchers from Yale submitted applications for a lab manager position to faculty members in the biology, chemistry, and physics departments at a number of research universities. The application materials were identical, except that half were assigned a female name, and the other half assigned a male name. Science faculty were asked to evaluate the applicants’ competence, hireability, and mentoring potential (how deserving they were of mentoring), and also to recommend a starting salary.
The results were dismaying, to say the least: the researcher report that

Faculty participants rated the male applicant as significantly more competent and hireable than the (identical) female applicant. These participants also selected a higher starting salary and offered more career mentoring to the male applicant

What's especially neat about the setup is that the research team made the fictitious recent undergraduate student applicant something less than an obvious genius.  For instance, though "Jennifer/John" had good grades, the information also pointed out that her/his GPA wasn't quite stellar and that she/he had withdrawn from one course.  In short, the evaluators were offered a type of an average academic applicant, not a clear female or male Einstein clone.   It is in these "borderline" cases where mentoring, for instance, matters greatly.

The fictitious resumes were sent to the faculty in appropriate departments in six large research-oriented universities, picked from various parts of the US.  The final study used the answers from 127 participants with these characteristics (pdf):

Of participants, 74% were male and 81% were White (specific ethnic backgrounds were reported as follows: 81% White, 6% East-Asian, 4% South- Asian, 2% Hispanic, 2% African-American, 2% multiracial, and 1% each for Southeast-Asian, Middle-Eastern, and other), with a mean age of 50.34 (SD = 12.60, range 29–78). Of importance, these demographics are representative of both the averages for the 23 sampled departments (demographic characteristics for the sampled departments were 78% male and 81% White, corresponding closely with the demographics of those who elected to participate), as well as national averages (9). Additionally, 18% of participants were Assistant Professors, 22% were Associate Professors, and 60% were full professors, with 40% Biologists, 32% Physicists, and 28% Chemists. No demographic variables were associated with participants’ substantive responses (all P > 0.53). As expected when using random assignment, participants’ demographic conditions did not vary across experimental conditions. Because there were 15 female and 48 male participants in the male student condition, and 18 female and 45 male participants in the female student condition, we obtained sufficient power to test our hypotheses (10).
Each participant got the fictitious resume only once, 63 from John and 64 from Jennifer.  Thus, the study does not (and cannot) measure any one particular faculty member's gender bias but the average bias that might appear when all the evaluations are judged together.

The respondents were asked to rate the competence and likability of Jennifer/John, to state whether she/he is worth hiring, to suggest a suitable starting salary and to indicate whether they themselves would be willing to mentor Jennifer/John.  The study supplement states that the participants were told Jennifer/John is a real person and that the evaluations would be sent to her/him.

What are the findings?

OOPS.  Here's a table from the study itself (pdf):

And here is a table (pdf) with information on how female and male faculty members judged the applicant:

In words, the same fictitious resume was judged to belong to a more competent applicant when the only change in it was from "Jennifer" to "John."  Jennifer was judged less worthy of hiring or mentoring and she would have been offered a lower starting salary.

The study also argues that "Jennifer" was judged more likable than "John."  But what seems to drive the results is the competence variable.  Being called "Jennifer" rather than "John" makes the applicant less competent!

Fascinating stuff.  But most people aren't fascinated by that most astonishing finding of all:  The Power In A Name.  Based on the comments attached to Geier's post on the study, many start immediately digging for reasons why the name difference should be powerful:  Women are more likely to bugger off because of pregnancies and gendered family duties, women are less likely to want to work 80-hour weeks and so on.

Of course all of that is statistical discrimination.  The two applicants give identical data and the judges then add their own views about women to the stew.  But what remains is still discriminatory, given that nobody knows if Jennifer/John would act according to the presumed averages or stereotypes of her/his gender.

Some have commented on the finding that female faculty members exhibit the same (possibly unacknowledged biases) as male faculty members.  I guess the idea is to point out that it's not really gender discrimination if women do it, too*?  Not sure, but of course we are all taught the same general views in the society and it would be extremely unlikely that women would have some sort of a general immunity to those views.

Or put in other terms, IF the perhaps unacknowledged biases are created in this case from, say, what one observes inside the science professions, then both female and male faculty members would notice that there are fewer women than men in the profession, that there are fewer women in the top ranks and that more women than men are lost in the pipeline.  The theory of statistical discrimination would then suggest that such "incomplete" data as the applicant's resume must be corrected downwards for Jennifer by the general likelihood that a randomly drawn woman (as opposed to a randomly drawn man) would succeed in the occupation, given that men seem to be doing better, as a group.  "John" doesn't require the same correction.  Indeed, in theory he might benefit from his group membership in the team "men."

Nevertheless, I found the mentoring results a bit discouraging.  Suppose that most of this bias indeed IS unacknowledged, meaning that the judges are unaware of anything sexist going on.  I would still have expected that female faculty members would have been more aware of the importance of mentoring women in their professions.  On the other hand, perhaps the judges answered the questions on the basis of "what's-in-it-for-me."  From that point of view the prediction that "Jennifer" would do less well than "John" means that mentoring the former would not "pay" for the faculty member the same way that mentoring the latter would.

I combed the study and the supporting material for a while yesterday,  in an attempt to find something to criticize in its methodology and so on.  It's possible that I missed stuff but mostly the research seems pretty good.  The one thing I would have liked to see is an analysis of the interaction terms between faculty member's gender and professorial level variable.

The supplementary material (pdf) notes that the researchers could not analyze the possible effects the institutions themselves might have on the results because doing so would have removed the privacy of the answers. But given that only six universities were included, it's not impossible that the faculty members discussed the study among themselves.  It's hard to know how something like that might have affected the findings.  The double-blind frame of the study could have been violated if two participants, say, found out that they had the same data but one had Jennifer and the other one had John.   On the other hand, if both of them happened to have a "John" or a "Jennifer", the discussion could have made their judgments more similar.

All that is trivial, I suspect.  The main take-home message from this study is that Names Matter in the science fields, or that there appears to exist a general statistical downwards correction, applied only to female applicants.  This takes the form of lower judged  competency and might affect starting salaries, getting hired and later mentoring.  Virginia Valian's drip-drip theory of discrimination then suggests that these minor differences in, say, starting salaries and the amount of mentoring might ultimately pile up into large cumulative differences in career successes.

None of this is intended to negate the role of alternative explanations for the dearth of women in the mathematics and science fields.  But the study does suggest that plain prejudice still bites.

*This needs a little more explanation:  The idea is that because female faculty members are women, too, they cannot be subtly biased about the class "women."  But humans are really good at viewing themselves differently from other people.  An example:  My children, the neighbors' brats.   And traditionally one way for ambitious women to cope with the gender bias is to assume the role of the exceptional or the role of the "honorary man."  When that role is assumed, being a rarity might, in fact, be a positive characteristic because it seems to validate how exceptional one is.

I didn't realize how common such thinking was before I began blogging and reading comments attached to various newspaper articles about gender.  Of course some of those misogynistic comments attributed to female nyms might not be by women.  But some probably are.  In short, we are all like the fish who think water has no taste because they live in it.  Similarly, we pretty much inherit the biases of our cultures, and, yes many of them swim below the surface thinking.