Saturday, November 22, 2014

Gender bias at SfN poster presentations

On the last day of SfN, before leaving for a plane, I was so tired that I could no longer think about science. I just could not force myself look at the posters and make a mental effort of understanding what is shown on them.

So instead, on the back side of my printed itinerary, I started counting how many spectators were standing in front of every poster, and also recorded the gender of the presenter. It was an instantaneous snapshot of one SfN poster session, and while it is not necessarily representative, I thought it could be fun to analyze offline.

Methodology:
  • SfN 2014, morning poster session of Wednesday, Nov 19 2014
  • For those rows I visited, I recorded number of spectators at every poster that had a presenter at the moment of my snapshot. While I skipped several rows, I covered posters from all parts of the hangar, from one wall to the other, so whatever poster themes were presented that morning, they should have all left a trace in my data.
  • "Spectators" were defined as people either listening to the presenter, or interacting with them in some meaningful manner in front of the poster.
Results and discussion:
  • A total of 749 poster presentations were observed. 
  • There were slightly fewer female (46%) than male (54%) poster presenters (p=0.01, binomial test assuming 1/1 split).
  • Average number of spectators in front of female-presented posters (1.68±1.59, n = 344) was slightly smaller than that in front of male-presented posters (2.08±1.90, n = 405; p = 0.002, t-test).
  • The distribution curve (above) suggests that the average numbers may differ for two reasons. While walking in the hall, I had an impression that most "lonely presenters" (N spectators == 0) were female, but actually this difference is not significant (p = 0.09, exact Fisher test). The "crowded" posters however were predominantly presented by male scientists (p = 0.01, exact Fisher test for N spectators > 5).
  • My impression is that the major reason for this skew is that at some point of their careers males and females self-selected to different parts of neuroscience, as gender ratio in different "poster sessions" (rows of posters) varied a lot (lower quartile = 38% female presenters ; higher quartile = 63% ; anova p = 0.01). Further analysis showed that different rows had different average number of spectators even after adjusted for varying female-to-male ratio (mixed model ancova F93,654 = 1.8, p = 1e-5). In fact, this effect was much stronger than the female-male bias, which should not be too surprising for any SfN attendee: different fields of neuroscience get very different attention. My guess therefore is that the difference across disciplines is primary, and then it "bleeds" into gender bias through self-selection.
  • But of course there may be biases at other levels as well: who gets to better labs, who gets better projects within every lab, etc. And on top of that, there may be a bias at the level of poster presentation itself.
  • Also the data may be confounded by the fact that 2nd and 3d world countries, that on average present less exciting posters, can also have very different ratios of female and male scientists, but I cannot control for it at this point. Pity, that.
Now I'm wondering if I ran ancova correctly; never tried it before...