For many early-career scientists, the National Science Foundation’s (NSF’s) graduate fellowship is the award to win. Officially called the Graduate Research Fellowship Program (GRFP), it provides its 2000 or so annual awardees with 3 years of funding. It can give students a leg up when applying to graduate programs or trying to earn a spot in a coveted lab. It also confers a certain cachet among colleagues and can jump-start a strong CV for future applications. “We’re trying to support the next generation of STEM leaders,” says Nirmala Kannankutty, acting director of NSF’s Division of Graduate Education, which manages the GRFP.
But those “STEM leaders” overwhelmingly come from the same few institutions, highlighting how inequality propagates through the academic system. “It’s another way the rich get richer,” says Matthew Cover, a professor of ecology at California State University (CSU) Stanislaus in Turlock, who has analyzed how recipients are distributed across institutions. As this year’s eligible students begin to labor over their applications, it’s worth asking why certain schools come away with dozens of winners each year—and what that means about the academy at large.
An uneven distribution
The GRFP is designed to support early-career graduate students in science, technology, engineering, and mathematics (STEM) fields. U.S. citizens who have not completed more than 12 months of a graduate degree program can apply. The application requires an original research proposal; short statements detailing personal experiences, background, and goals; academic transcripts; and three reference letters. Awardees’ funding pays on the high end of what a standard research assistantship or teaching assistantship stipend would net, and they receive access to GRFP programs such as international research collaborations and career development internships.
But this opportunity is not evenly distributed. According to Science Careers's analysis of this year’s recipients, the 10 schools with the most grad student awardees amassed 31% of all the grants. And 14% of recipients are at the top three alone: the University of California (UC), Berkeley; the Massachusetts Institute of Technology (MIT); and Stanford University. More broadly, in 2017 86% of awards went to students at R1 (very high research activity) institutions, according to Cover’s analysis. Just 0.3% went to historically black colleges and universities, and none went to tribal colleges or universities.
A similar trend holds among awardees’ undergraduate institutions, according to an analysis of GRFP winners between 2011 and 2018 by statistical geneticist Natalie Telis, a former Stanford Ph.D. student who publishes data-driven analyses of social issues on her blog. (She now works in R&D at a genomics company.) Three schools—UC Berkeley, MIT, and Cornell University—swept up 10% of the awards. The top 10 schools accounted for about 25% of awards, and the top 30 made up about 50% of the entire pool. Winning a GRFP award is “a feather in the cap, rather than a transformative resource,” Telis says.
The most obvious explanation for these disparities is that more students from these universities apply. Currently, we can’t empirically evaluate this theory; NSF does not release information about applicants, and top schools—including UC Berkeley, Stanford, Cornell, and MIT—did not provide data about their applicants in response to Science Careers's request. But there is reason to believe that top-awardee schools have the resources to help their students craft competitive GRFP applications—benefiting both the students and the schools.
Full disclosure: I’m a product of this system. Both my undergrad and grad institutions are among the top producers of GRFP awardees, and I received an award when I applied in my first year of grad school.
Within the first week of my graduate program at UC Berkeley, it became obvious that all eligible students were expected to apply. Each weekly department newsletter included information on applying and reminders of the deadline. Our department’s director of student services scheduled a workshop for the 13 students eligible to apply, providing pizza as an incentive for attendance and inviting previous winners from the department to give tips. Students could opt to be connected with a mentor to review our applications, and we had access to a binder full of laminated applications from previous winners, as well as access to documents and still more applications through the department’s online portal.
I’m not certain how many of us ended up applying, but I know that four people received awards. That means at the most conservative estimate, our department’s award rate was about 30%—double the overall GRFP success rate of about 15%, calculated based on rough applicant numbers available for 2011 to 2018.
My experience isn’t unique among well-resourced institutions. According to 2018 GRFP recipient Jason Chang, a biomedical engineering graduate student at Cornell, the GRFP was “emphasized from the beginning in my program’s required first-year seminar course,” where students completed prompts and exercises to prepare them for crafting their applications. An awardee who studied astronomy at Columbia University said they attended a campus essay feedback workshop with GRFP recipients and had the opportunity to meet with a writing consultant for help with their personal statement.
More broadly, 18 of the 20 schools with the most GRFP awardees are also in NSF’s top 50 schools for R&D expenditures, and NSF’s top 10 R&D schools generated 20% of this year’s awardees. Telis’s analysis also suggests resources play a role: The most expensive undergraduate institutions tend to have a glut of awardees.
Smaller or less well-funded schools, in contrast, may not have administrators to send out application deadline reminder emails, writing tutors available for essay workshopping, or the funds for pizza-fueled advice panels. At CSU Dominguez Hills, for instance, professor Terry McGlynn says his biology department has just 10 tenure-track faculty members for 600 undergraduate students. “A good number of students are interested in going to grad school, but we can’t sit down with all of them and work on the applications,” he says.
That’s in line with Janessa Stewart’s* experience. She applied for the GRFP last year while completing an undergraduate degree at a West Coast community college, but there were “absolutely no resources” to help her, she says. “In fact, I found out about [the] NSF GRFP through science Twitter.” Other scientists on Twitter shared winning applications and offered to read Stewart’s application, which helped. But she did not receive the award, which complicated her plans for graduate school. She had a standing offer for a spot in a program, but she needed a faculty member to fund her work, as well as funds to help her move to attend the program. She found ways to make the financial side work—she secured an adviser’s commitment for funding and crowdfunded to help make the move—but the “GRFP would definitely have eased my way into grad school,” she says. She plans to apply again this fall. (Her name has been changed at her request due to concern that her candor could affect her chances of winning.)
To help students at less well-resourced schools, like Stewart, McGlynn has a proposal pending with NSF to build a cross-institutional mentoring program, which would match students with mentors across schools. The hope is that such opportunities would support applicants in crafting competitive GRFP applications, giving them a better shot at the award. But students at these institutions will likely still face additional challenges.
For one thing, students with family responsibilities, a job, a disability, or other responsibility or barriers may not have the time or energy to devote to tackling an application. “The GRFP is long and complicated and requires real time and commitment to draft—people need affirmation that it’s going to be worth their time,” says Annalisa Price*, a professor at a midsized private university who has reviewed GRFP applications. (She has requested that she be referred to by a pseudonym because NSF reviewers are asked not to discuss the review process.) “So many people might talk themselves out of it,” she says—including those who could benefit the most.
Who’s destined for greatness?
There may also be bias on the other side of the application: among the reviewers. NSF mandates that all GRFP reviewers—science researchers and graduate education experts the agency selects—undergo bias training. But, as Kannankutty notes, “in the case of the [GRFP], where we’re actually assessing individuals as opposed to research, the propensity for bias is even greater.” Determinations of whether early-career applicants have potential as “leaders” may be especially prone to snap judgments. After all, students at the cusp of their careers have a limited track record in research, so reviewers may default to looking for other markers of success—and that’s where bias can sneak in.
“When the NSF says, ‘Our goal is to identify students who have potential to be high achieving scientists,’ I think there’s a lot of assumptions about what that path looks like,” Cover says—namely, completing a bachelor’s degree at a top university and immediately beginning a graduate program at another prestigious institution. Plus, “people’s own experiences certainly color how they think about this issue of what is an academic pathway that shows someone has the potential to be a great scientist.” A professor who was trained at top R1 universities and graduated with a Ph.D. at 27—a common trajectory among faculty members—may be more likely to believe that others who have followed similar paths are the type of “STEM leader” the GRFP seeks to support.
“In theory this is supposed to be about promise,” Price says. “But in practice, I think it’s the Matthew Effect,” referring to how work from well-known scientists receives greater recognition. “You’re going to think the most promising person is the person who already had access to the most resources and support.”
That’s the type of feedback that Stewart received in her rejection. The reviewers “talk about me falling short compared to my ‘cohort,’ but unless they’re comparing me to other low-income disabled postbaccalaureate students at community colleges, they’re not comparing me to my true cohort,” she says. “I think I’m precisely the sort of student the GRFP should be supporting.”
In 2016, NSF announced graduate students would only be eligible to apply for the GRFP once during their graduate career; before, applicants had two opportunities. The idea, it seems, is that undergraduates or postbaccalaureate applicants—“a more diverse population than admitted graduate students,” as the announcement puts it—would stand a better chance at winning an award if they weren’t competing against second-year graduate students, and those improved odds might encourage more undergrads to apply.
It may be too early to determine the effects of this change, but Telis’s analysis isn’t promising. After the switch, the top 10 institutions attended by awardees as undergraduates garnered more winners than before, and the overall number of undergraduate winners remained the same.
Perhaps it’s time to consider more radical solutions. On Twitter, McGlynn asked whether a cap on the number of applicants per school may help level the playing field. Universities would vet applicants and decide on a certain number to advance to NSF, giving less-represented institutions a better shot at winners. But McGlynn acknowledged critics who pointed out that such a process might just introduce bias at the institutional level instead.
Whatever NSF and the community do, the key is to acknowledge inequity, analyze it, and work toward a solution. “These questions are something we have a responsibility to ask as a scientific community every year,” Telis says.
* Names have been changed.