I had a thought occur to me over the past few days. It's been growing along at the back of my mind and is only partially crystallized.
What if PIs of a given class of interest, whether that be sex, ethnicity, nation of origin or whatever, are not randomly distributed across the various topic domains supported by the NIH? What if a PI of characteristic X tends to work on Topic B using Model M whereas a PI of characteristic Y tends to work on Topic A using Model H?
What if the funding rates for Topic X differed from those for Topic Y? Or if applications using Model M consistently succeeded differently compared with applications using Model H?
I didn't see any covariates for topic domain or even the funding IC in the Ginther report.
Surely someone at NIH is thinking about this. Surely?
I have two anecdotes for your consideration.
First, as with many areas of science, the ones dear to me suffer from a sex bias. There is a huge tendency to do the animal studies in male animals. Any study using female animals is very frequently a sex comparison study and is proposed explicitly or implicitly as a comparison with the default, i.e. male. I've talked about this before. The NIH also takes pains to fix the generalized reluctance via their most functional technique, the call for applications for a dedicated pool of money. In theory, the awarding of grants on sex-differences or on issues specific to women's health will then spur additional work. Perhaps create a sustained program or even a career of work on this topic.
My anecdote is that I've noticed over the years (possible confimation bias here) that women in my field have a greater representation than men in these sorts of studies. Sex-differences models and womans' health issues in my fields of interest seem to have women as the driving investigators more often than their overall representation.
If this generalizes, then we will want to know if the competitive success of such grant applications because of topic is contaminating our estimation of women PI's success.
The second anecdote is older and comes from my long history participating on the "Diversity" committees of various academic institutions. Back in the dark ages I recall an incident where a Prof in the experimental sciences had to go to war with a Dean who was in charge of undergraduate summer research funds for underrepresented individuals. The Prof had a candidate who wanted to work in the experimental science, but the awards were generally being made to kids who wanted to work on academic topics related to underrepresented groups. The Dean thought this was the most important thing to do. In this case the prof won his battle in the second year of trying, over the objections of the Dean. I keep in touch with some of my undergraduate professors and I can say that said undergrad went on to become a NIH funded investigator (who still fails to work on issues directly related to underrepresentation). I have no idea if any of the other underrepresented summer research students went on to glorious academic careers in their respective disciplines, perhaps they did. But this is not the point. The point is that perhaps I am a little too glib about the pipeline implications of Ginther. Perhaps the grooming of underrepresented minority undergrads for a career in academics is itself not topic neutral. And the shaping and shifting from that very early stage may dictate field of study and therefore the eventual success rate at the NIH game.
Assuming, of course, that Topic X enjoys differential success rate from Topic Y when the grants are under review at the NIH.