Expanding Science Fellowship Opportunities
Federal agencies must do more to ensure that prestigious fellowships for science graduate training are awarded in a manner consistent with larger goals of equity.
Famed astronomer Amy Manizer, physicist Eric Cornell, economist Steven Levitt, and Google cofounder Sergey Brin all have at least one thing in common: early in their careers, they received fellowships from the prestigious National Science Foundation (NSF) Graduate Research Fellowship Program (GRFP), which, in the words of the program’s homepage, “recognizes and supports outstanding graduate students.” Since the program was established 70 years ago, more than 60,000 GRFP fellowships have been funded, including 42 Nobel Laureates and more than 450 members of the National Academy of Sciences. Receiving a prestigious fellowship can aid the recipient’s path to completing a graduate degree. It can also serve as a springboard to a successful and productive career. According to the GRFP website, “the reputation of the GRFP follows recipients and often helps them become life-long leaders.”
Although it is well-known that the majority of doctoral students (around 81%, according to the most recent data) are trained at the highest-level research universities as categorized by the Carnegie Classification (known as “R1 universities”), GRFP awards go even more disproportionately to students at these institutions. A 2014 evaluation of the GRFP program revealed that, of PhD completers, nearly 95% of GRFP recipients attended R1 institutions. In more recent years, the distribution of GRFP fellowships across institutional types has remained unequal, with 31% of fellowships going to students at just 10 institutions.
There may be various reasons for this distribution, including the prevalence of certain fields of study at different types of institutions and the accessibility of those fields to students from different backgrounds. But GRFP and other prestigious fellowships are designed primarily as awards for individuals judged to have excellent potential, not for specific research projects (although a research proposal is part of the application). Such talented students will be found at all types of institutions. We suspect that one reason such fellowships are awarded disproportionately to students at R1 universities is because those institutions tend to have more resources and may do more to help prepare students to apply. However, “suspect” is as far as we can go: even basic data are lacking.
The disproportionate number of prestigious graduate fellowships awarded to students at a small number of universities is an overlooked part of a well-known problem. Prominent voices in the science policy community have spoken out about the need to increase the geographic and demographic diversity of the science, technology, engineering, and mathematics (STEM) workforce. Leaders have put forth concepts such as “missing millions” or “lost Einsteins” to describe those who have been excluded from STEM. More equitable geographic distribution of research funding could promote local economic and health benefits, and has recently garnered increased congressional attention.
Although federal fellowship programs like GRFP may be relatively small in the number of students they fund—most federally supported graduate students in STEM receive research assistantships—they loom large in the culture of science. Thus, they tend to reinforce a narrative that scientific excellence is defined by the school a graduate student attends, their adviser’s prestige, or the level of access they have to grant writing workshops and other resources, which perpetuates an inequitable system that overlooks many qualified applicants. Agencies such as NSF must do more to make certain their fellowship programs are serving the larger goal of STEM equity, including collecting and reporting better data, ensuring that applicants at all institutions have training resources, and experimenting with other ways to equalize the fellowship process.
To better analyze fellowship programs and determine how to make them more equitable, federal agencies must work harder to collect data. In researching these fellowship programs, we were repeatedly stymied by a lack of data. Although the GRFP makes available a list of all awardees and honorable mentions by institution and field, data on applicants are lacking. With success rates of 12–16% over recent years, it is important to have a better understanding of the full applicant pool in comparison to the awardees. For example, one proposal to make GRFP awards fairer would involve limiting the total number of applications per institution. However, assessing the utility of such a proposal is difficult without data. The most recent full evaluation of the GRFP program that we are aware of is from 2014. Beyond making available more complete data about the applicant and reviewer pools—including, to the extent practicable, demographic characteristics, institutional characteristics, and geography—we believe the program should undergo more regular evaluation, perhaps every two or three years, and that this information should be made publicly available.
A second area that the NSF should address is providing guidance to applicants. The R1 universities are classified as such in part because of their robust research infrastructure, and students from these institutions who wish to apply for fellowships often have special resources at their disposal. Some research universities offer fellowship workshops or boot camps for students and may provide additional financial incentives for students to seek fellowships from sources outside the institution. Many such institutions have fellowship recipients and experienced advisers on campus available to help graduate students with the application process. For example, Columbia University, a private R1 institution, has a multitude of grant writing support systems, including a semester-long grant writing course, workshops held several times a year specifically for the NSF GRFP and National Institutes of Health (NIH) F31 fellowship applications, access to personal review of application material by experts, and peer support systems.
In contrast, in the experience of one of us (Damico), the University of North Carolina at Greensboro (a public R2 or “high research activity” doctoral university, according to the Carnegie Classification) provides no grant workshops, support groups, or even announcements about graduate fellowships, including the NSF GRFP. While these examples are anecdotal, they paint a picture of inequity in institutional grant writing support that is likely true more broadly. Because this issue is not well documented, it remains a mystery where exactly the need lies for more support at other institutions and how successful support systems can be applied to meet the needs of students at other universities.
We strongly encourage NSF and other agencies with selective graduate fellowship programs to investigate disparities in grant writing support and take steps to increase access to quality support across institutions. Data indicate just how helpfulthese programs can be to funding success for scientists of all levels, including graduate students. For example, a studyfrom the University of Maryland, Baltimore County, surveyed students before and after a mentoring program that included a workshop series designed to inform them about best practices when applying to the GRFP. The survey data showed that participation strongly improved their knowledge of the process and expectations, as well as their confidence in being able to write a successful fellowship application after the program. Federal agencies could design and disseminate their own grant writing programming to students interested in applying to these fellowships, which could fill the gaps at institutions that are unable to implement their own programming and help level the playing field for all applicants. Providing this information in a virtual and free format would also help alleviate accessibility issues due to money, time, and travel resources.
Beyond collecting and releasing more data and providing application preparation training and support, agencies should experiment with other ways to make the fellowship process more equitable. Two specific ideas include adding a “funded pending revision” decision category to fellowship award decisions and taking steps to anonymize reviews.
One problem for students who lack institutional support is that their fellowship applications may be rejected with little feedback and there is often no chance to submit revisions until the next application cycle. For applicants to the GRFP, the stakes are even higher: beginning in 2017, the program changed its rules to allow students to apply only once. A change that could ease some of these burdens would be to establish a new category for competitive proposals, perhaps called “funded pending revisions,” that allows some applicants the opportunity to incorporate reviewer feedback to strengthen their submission. A similar concept, “accepted pending revisions,” is used by some scientific journals. Such a policy could both broaden the awardee pool and allow students to learn from the process.
Another area for investigation is reforming the GRFP review and selection system—a process that has been described as “dysfunctional”—to create clearer standards for judges, track accountability, and ensure the diversity of reviewers. For example, both NIH and NSF have experimented with anonymizing peer review, but should do additional work to understand the role of bias in review and learn whether anonymizing can help further diversify the pool of awardees in their prestigious fellowship programs. Along with anonymizing the process, agencies should continue to take steps to broaden the reviewer pool, specifically including reviewers from a variety of institutional types, which could also help in building support programs at these institutions. The NIH Center for Scientific Review has recently committed to “diversifying and broadening” its reviewer pool and will be regularly updating data on reviewer demographics. NSF is also exploring how best to attract “an untapped reviewer pool” and remove “barriers to scientists serving as reviewers,” according to an interagency policy group working on the issue. For the GRFP specifically, NSF accepts volunteer reviewers, which could affect the types of applicants selected. More research to understand the factors that make individuals more or less likely to review could help broaden the pool of reviewers. In addition, as we suggested earlier, data on the reviewer pool should be reported and made part of regular program evaluation.
Pursuing a doctoral degree is a difficult endeavor for many reasons. A recent brief from the National Science Board, for instance, describes financial barriers that hinder doctoral students, especially those from first-generation and lower-income backgrounds. Prestigious fellowships are one way that federal agencies can ease the way for promising students and increase their potential for longer-term career benefits. At a time when so much attention is being paid to STEM equity, differences between institutions in research infrastructure and resources should not be what makes one student more excellent than another. Who knows how many more future STEM leaders are out there, hoping the federal government can help unleash their full potential?