Is it Seniority Bias or Gender Bias?

Feb 19 2008 Published by under NIH Budgets and Economics

PhysioProf was recently discussing a figure from the NIH Extramural Data Book which pointed to a poorer success rate for women submitting competing continuation applications but not for new projects. His title "Old Boys' Network Favors Men's Continuing Grants?" was perhaps intentionally rich since as I pointed out in the comments, these results could be explained more by the "Old" part of the descriptor than by the "Boy's" part. Additional data from the NIH Data Book provide more support for the seniority hypothesis.


Starting with the figure comparing the percentage of all awards that were to women PIs across "Career" (e.g., K mech), "Research Project Grants" (R mech) and "Centers" we find that while women are clearly underrepresented across the board, the severity of the effect is correlated with relative seniority of the mechanism, if you will permit me. Since I can't make these very legible the y-axis runs 0-40% in 5% increments for data from 1998-2007 Fiscal Years.

The impression of a bias for the RealMoney (i.e., Centers and "other", read contracts and U mechs) to accumulate in the hands of the males (light blue bar) is reinforced by another slide. The right y-axis runs to $2.5 million and the left y-axis is 0-100% in 10% increments representing the average of award $$ to women divided by the amount awarded to men PIs. So yes, women who are awarded a RPG get on average 90% of the money and those awarded Centers get 70% of the money. [Sidebar on my infuriation with NIH slides like this, does this award number include indirect costs? The average dollars for RPGs would argue that it does. So yet another factor would be the degree of gender bias at high-indirect-cost institutions vs. low-normal indirect-cost institutions. Argh, these data start to look nearly meaningless as you get into them!] For those of you that think that Centers may be a bit of a boondoggle, here's another critique!

One further interesting thing I noticed was the slide comparing women's participation on study section (yellow) versus the applicant pool (green). As readers are aware, I'm focused on the effects on the study section in setting the stage for grant funding because this is one of the most important stages of decision making. This figure shows that women are proportionally overrepresented on study section in comparison with their participation in the applicant pool. This is a GoodThing. It surprised me a little because going by my subjective impression I thought women were even more overrepresented on my section. Going back to check, I find that we have 42% women on the standing roster (and 39% women overall at the latest review) so I guess we do a little better than CSR average. These data address two slightly contradictory points, however. Diversity of the study section does not automatically confer group advantage even to those who are overrepresented. Thus, attempts to increase representation from ReallyDismal to SlightlyBetter (for transitioning and junior scientists, for example) are unlikely to cause wholesale reorganization of the review outcome. You may read this as an argument that we need to go from ReallyDismal to QuiteGoodActually or a defense that SlightlyBetter isn't going to end the world AsWeKnowIt, depending on where you sit.

To wrap up, my point here is not to minimize any impression that women are less well represented in the NIH grant game than are men as PIs. The data present a compelling picture that this is the case. It is important, however, to dissect the source of bias leading to this reality. One can see quite clearly that the remedies for seniority-bias, mechanism-bias and across-the-board gender-bias should be different. And if all three are contributing factors, well, then multiple remedies are required. Furthermore, the data remind us that our hypotheses are testable. So if the goal in ensuring that women were represented on study section closer to the general population than to the pool of applicant scientists was to increase women PIs closer to the general population level, well, we can see how that is working out.

11 responses so far

  • seed42 says:

    women who are up top are pulling up the ladder, so to speak

  • DrugMonkey says:

    women who are up top are pulling up the ladder, so to speak
    umm, no. that's not really what I would make of this. that would imply that woman reviewers are hammering female PIs worse than male reviewers are and there is no evidence for that here. And while they most certainly have this data inside CSR I doubt very much that a breakdown of how GroupX reviews applications from GroupY is ever going to see the light of day...

  • Becca says:

    Would you expect a senority bias to decrease with time? Yes, the pipeline is incredibly leaky (as the data make clear), but isn't the percentage of female full professors increasing (very) gradually over time?

  • DrugMonkey says:

    but isn't the percentage of female full professors increasing (very) gradually over time?

    This is the theory.
    Similar to the one that suggests that GenX-ers are the generation that is so steeped in Women's Lib, ERA, women in all of the workplaces, etc that we will experience a radical shift in attitude once we get rid of the bad old Boomers.
    Color me impatient.
    And cynical. People have a way of getting more conservative as they age, of coming to believe memes that work in their favor and of justifying tremendously self serving perspectives on life...

  • whimple says:

    Seniority Bias or Gender Bias?
    Not to get all un-PC or anything, but has it been factually established that the differences in aggregate statistical success rates are actually due to bias?

  • DrugMonkey says:

    but has it been factually established that the differences in aggregate statistical success rates are actually due to bias?
    there are several ways to think about "bias".
    in the descriptive sense, a difference from expected value is sometimes referred to as a "bias". The data PP presented show a difference between male and female PI "success" on competing continuations so the descriptive sense is verified.
    My discussion is trying to get at what I believe you are asking. Is this difference attributable to a sort of aggregate of case-by-case unfair treatment of woman PIs apps? or is this an artifact of collapsing across all stages of continuation? If the latter, this pushes the discussion of "bias" into questions of why women are less well represented as PIs of longer-duration continued projects and away from the question of acute individual review bias.
    Now, the NIH (and Congressional staffers) might not care about causation, if all their attention is placed on getting the curves to overlap. Me, I care about the source of the difference.

  • whimple says:

    No, what I'm asking is whether the senior men have higher success rates because in the aggregate, they are better scientists.

  • PhysioProf says:

    As I have pointed out before, male privilege is probably a much more important cause of gender inequity than "bias" against women. As I also pointed out, however, it is much easier for the priviliged to "root out bias" than it is for them to recognize and agree to give up their privilege.

  • DrugMonkey says:

    No, what I'm asking is whether the senior men have higher success rates because in the aggregate, they are better scientists.
    Define "better scientist" in a way that is objective and not circular and I'll take a stab at it...

  • whimple says:

    Define "better scientist" in a way that is objective and not circular and I'll take a stab at it...
    How about:
    Publishes data that provides deeper insights into more important issues.

  • DrugMonkey says:

    Hmm, can't believe I forgot to respond to this
    How about:
    Publishes data that provides deeper insights into more important issues.

    The answer to your question then, is "no".

Leave a Reply