Meta Research Awards

Meta Research Awards is sponsored by National Science Foundation (NSF). Supports research projects that advance the understanding of social media platforms, including Meta's services.

Meta-Research: Systemic racial disparities in funding rates at the National Science Foundation | eLife Meta-Research: Systemic racial disparities in funding rates at the National Science Foundation https://doi. org/10. 7554/eLife.

83071 Article and author information Concerns about systemic racism at academic and research institutions have increased over the past decade. Here, we investigate data from the National Science Foundation (NSF), a major funder of research in the United States, and find evidence for pervasive racial disparities. In particular, white principal investigators (PIs) are consistently funded at higher rates than most non-white PIs.

Funding rates for white PIs have also been increasing relative to annual overall rates with time. Moreover, disparities occur across all disciplinary directorates within the NSF and are greater for research proposals. The distributions of average external review scores also exhibit systematic offsets based on PI race.

Similar patterns have been described in other research funding bodies, suggesting that racial disparities are widespread. The prevalence and persistence of these racial disparities in funding have cascading impacts that perpetuate a cumulative advantage to white PIs across all of science, technology, engineering, and mathematics.

Federal science agencies steer and implement national research priorities through grant-making activities, administering funds to hundreds of thousands of researchers at colleges, universities, research institutions, and other organizations across the nation.

In 2019, more than half of all research expenditures at US higher education institutions were supported by the federal government through agencies like the National Science Foundation (NSF; $5. 3 billion) and the National Institutes of Health (NIH; $24. 4 billion) ( National Science Board, 2021b ).

As mainstays of the scientific enterprise in the US, these funding bodies exert major influence on the programs and priorities of all higher education and research organizations, conferring economic stability and social capital to individuals and institutions by awarding grants. In academia, particularly at research-intensive universities, grants underpin every aspect of a researcher’s capacity to produce knowledge and innovations.

Support for equipment and facilities, stipends and salaries for trainees and personnel, and publication costs all generally depend on grant funding. More funding leads to more research, publications, and reputational prestige, which attracts more talent and generates more research output. Thus, grant awards play a crucial role in research productivity and, by extension, the success and longevity of academic careers.

Funding agencies solicit proposals from principal investigators (PIs) and process them through an evaluation system, making awards based on scientific merit and potential benefit to society. However, several studies over the past decade have revealed inequalities in the allocation of research funding, most notably at the NIH. A 2011 study showed that Black PIs were funded at roughly half the rate as white PIs ( Ginther et al.

, 2011 ). Subsequent analyses surfaced additional inequalities across race ( Ginther et al. , 2012 ; Hoppe et al.

, 2019 ; Erosheva et al. , 2020 ; Lauer et al. , 2021 ; Ginther et al.

, 2018 ; Ginther et al. , 2016 ; Nikaj et al. , 2018 ), gender ( Ginther et al.

, 2016 ; Nikaj et al. , 2018 ; Oliveira et al. , 2019 ), age ( Levitt and Levitt, 2017 ), and institution ( Ginther et al.

, 2012 ; Hoppe et al. , 2019 ; Wahls, 2019 ; Katz and Matter, 2020 ). Despite a decade of efforts within NIH to reduce disparities, many remain or have worsened ( Taffe and Gilpin, 2021 ; Lauer and Roychowdhury, 2021 ).

Such findings are not confined to federal funding bodies: the Wellcome Trust, one of the largest philanthropic funders of scientific research in the world, recently identified similar disparities by race in the distribution of their awards ( Wellcome Trust, 2021 ; Wellcome Trust, 2022 ; Wild, 2022 ).

Here we ask whether racial funding disparities are observed at the NSF, the flagship US agency for science, technology, engineering, and mathematics (STEM) research. In contrast to agencies with mission-oriented priorities in biomedicine, space, and energy, NSF has the federal responsibility to support basic research in all areas of STEM, as well as STEM education and workforce development.

We examine data on funding rates, award types, and proposal review scores disaggregated by PI race and ethnicity from 1996 to 2019. These data are publicly available in federally mandated annual reports on the NSF merit review process and describe award or decline decisions for over 1 million proposals.

Demographic information is collected at the time of proposal submission, when PIs voluntarily provide information on ethnicity — Hispanic or Latino, or not — and race — American Indian or Alaska Native (AI/AN), Asian, Black or African American (Black/AA), Native Hawaiian or other Pacific Islander (NH/PI), and/or white.

Because the contents of the merit review reports have evolved over the years, some data are only available for a limited period (e.g., review scores are only available for 2015 and 2016). Nevertheless, we examine all available data to describe and analyze patterns in funding outcomes by PI race and ethnicity. The NSF receives tens of thousands of high-quality submissions each year, many more than it can fund ( Figure 1A ).

From 1996 to 2019, the overall funding rate, or the proportion of proposals that were awarded, fluctuated between 22% and 34% due to factors such as changing budgets and proposal submission numbers. For example, stimulus funding in 2009 from the American Recovery and Reinvestment Act raised the funding rate to 32%, whereas a significant increase in proposals in 2010 lowered the funding rate to 23%.

Figure 1 with 3 supplements see all From 1999 to 2019, proposals by white PIs were consistently funded at rates above the overall average, while proposals by most other groups were funded at rates below the overall average. ( A ) Overall funding rates (black line) and total number of proposals (gray bars) have fluctuated on a yearly basis over time. ( B ) Racial disparities in funding rates have persisted for more than 20 years.

Funding rates by PI race and ethnicity are normalized to the overall rate for each year. Groups represented by thinner lines submitted on average fewer than 500 proposals annually. Data for white and Asian PIs are only available starting in 1999, and for multiracial PIs starting in 2005.

Source data: Data S1 in the accompanying data repository ( https://doi. org/10. 5061/dryad.

2fqz612rt ). Despite year-to-year variability in the overall funding rate, there are persistent and significant differences in funding rates between proposals submitted by PIs from each racial and ethnic group. To investigate these differences, for each year, we calculated relative funding rates for each group, normalizing by the annual overall funding rate.

We find that proposals by white PIs were consistently funded at rates higher than the overall rate, with an average relative funding rate of +8. 5% from 1999 to 2019 ( Figure 1B ; Figure 1—figure supplement 1 ). The relative funding rate for proposals by white PIs also steadily increased during this period, from +2.

8% in 1999 to +14. 3% in 2019. In contrast, proposals by most non-white PIs, specifically Asian, Black/AA, and NH/PI PIs, were consistently funded below the overall rate, with average relative funding rates of –21.

2%, –8. 1%, and –11. 3%, respectively.

The relative contributions of proposals by PIs from each group remain unchanged despite shifts in the number of proposals submitted by each group over time. Submissions by white PIs comprise the majority of proposals throughout the study period ( Figure 1—figure supplement 2 ).

In 2019, the competitive pool of proposals included 20,400 submissions by white PIs (66% among proposals from PIs who identified their race); 9,241 by Asian PIs (29%); 1,549 by Hispanic or Latino PIs (5%); 929 by Black/AA PIs (3%); 99 by AI/AN PIs (0. 3%); and 47 by NH/PI PIs (0. 2%) ( Figure 2 ).

Groups with fewer proposals experienced the greatest year-to-year variability in relative funding rates. In 2019, racial disparities in funding rates corresponded to hundreds of awards in surplus to white PIs and hundreds of awards in deficit to other groups. Each box represents 10 proposals.

Light gray boxes are unsuccessful proposals; colored boxes are funded proposals (awards). The black outlines represent 27. 4% of the proposals submitted by each group, where 27.

4% is the overall funding rate in 2019. For each group, the number of awards above (surplus) or below (deficit) this threshold is in bold. This graphic does not include proposals by multiracial PIs or PIs who did not provide their race or ethnicity.

Source data: Data S1 in the accompanying data repository ( https://doi. org/10. 5061/dryad.

2fqz612rt ). These persistent funding rate disparities are realized as large differences in the absolute number of proposals awarded to PIs in each group. For example, of the 41,024 proposals considered in 2019, the NSF selected 11,243 for funding, or 27.

4%. Proposals by white PIs were funded above this overall rate at 31. 3%, yielding 6,389 awards ( Figure 2 ).

If proposals by white PIs had been funded instead at the overall rate of 27. 4%, only 5,591 proposals would have been awarded. Thus, an “award surplus” of 798 awards was made to white PIs above the overall funding rate in 2019.

In contrast, proposals submitted by the next largest racial group, Asian PIs, were funded at a 22. 7% rate, yielding 2,073 awards. If the funding rate for proposals by Asian PIs had been equal to the overall rate, one would instead expect 2,505 awards, or 432 additional awards.

We refer to the number of awards required to bridge such gaps in funding rate as the “award deficit. ” Research awards are the standard mechanism through which the NSF funds PIs and institutions, comprising 71–76% of all awards from 2013 to 2019 ( Figure 3B ).

The remaining 24–29% of awards consist of grants for other activities and expenses, such as exploratory or early concept work; education and training; equipment, instrumentation, conferences, and symposia; and operation costs for facilities. The funding rate for these types of proposals, categorized by NSF as “Non-Research,” is generally 1. 4–1.

9 times higher than that for “Research” proposals ( Figure 3A ). Figure 3 with 1 supplement see all From 2013 to 2019, racial funding disparities were even greater for Research proposals, contributing to racial stratification in Research versus Non-Research activities. ( A ) Overall funding rates for Research proposals (dark dashed line) are more competitive than overall funding rates for Non-Research proposals (light dotted line).

( B ) 80–84% of all proposals and 71–76% of all awards were for Research activities. ( C ) Both Research and Non-Research proposals by white PIs were funded above overall rates. In contrast, Research proposals by PIs of most other groups were funded below overall rates, and at rates generally lower than those for Non-Research.

( D ) Only 46–63% of all awards to Black/AA PIs were for Research, far below overall proportions of awards for Research for all groups combined (black horizontal lines; panel B), contributing to a stratification of awarded activities by race. White text denotes the number of Research proposals or awards for each group per year; asterisk indicates that numbers are estimates based on available data.

Non-Research data for AI/AN and NH/PI PIs were not available 2017–2019. Source data: Data S2–3 in the accompanying data repository ( https://doi. org/10.

5061/dryad. 2fqz612rt ). When we examine racial disparities in the context of Research and Non-Research proposals from 2013 to 2019 (the years with available data), we find that disparities for Research proposals are generally larger ( Figure 3C ).

White PIs were the only group whose Research and Non-Research proposals were consistently funded above overall rates. The relative funding rates for Research and Non-Research proposals by white PIs also gradually increased, from +9. 0% and +5.

6% in 2013 to +14. 8% and +13. 2% in 2019, respectively.

In addition, for most years, Research proposals by white PIs had higher relative funding rates than Non-Research proposals. In contrast, Research proposals by PIs from nearly every other racial and ethnic group had negative relative funding rates, and were generally funded at lower rates compared to Non-Research proposals ( Figure 3C ).

In particular, for Black/AA PIs, the relative funding rates for Research proposals in 2013 and 2014 were anomalously low, at –35. 2% and –38. 9%.

These low funding rates meant that Research proposals by white PIs were funded 1. 7 and 1. 8 times more than those by Black/AA PIs in these years, with relative funding rates of +9.

0% and +10. 6% (absolute funding rates of 21. 3% and 22.

6% white versus 12. 6% and 12. 4% Black/AA).

For Asian PIs, relative funding rates for Research proposals fluctuated between –24. 0% and –14. 2%, for an average of –19.

1% from 2013 to 2019. Whereas the relative Research proposal funding rate for Black/AA PIs gradually increased to –9. 9% in 2019, the rate for Asian PIs did not.

Similar or worse outcomes are observed for Research proposals by NH/PI PIs, especially in 2015, when only 1 of 23 Research proposals were awarded (4. 3%). These Research funding rate disparities contribute to a stratification in awarded activities by race: from 2013 to 2019, only 46–63% of awards to Black/AA PIs were for Research.

This percentage is far below the proportion of Research awards to white PIs over the same period, 70–77% ( Figure 1D ). Although part of this stratification can be attributed to Black/AA PIs submitting proportionately more Non-Research proposals, these proposals were still consistently funded less often compared to Non-Research proposals by white PIs.

These results show that larger racial disparities in Research proposal funding rates are masked within the funding rates for all proposals. For example, in 2013, the difference in relative funding rate for Research proposals versus all proposals for Black/AA PIs is large, –35. 2% for Research compared to –18.

3% for all proposals ( Figures 1B and 3C ). In addition, these results suggest that the magnitude of disparities for Asian and Black/AA PIs is more similar when considering only Research proposals: on average for the 2013–2019 period, Research proposals by white PIs had a 1. 37- and 1.

40-fold funding rate advantage over those by Asian and Black/AA PIs, respectively (Data S14). Thus, disaggregating funding statistics by proposal type reveals a more complete picture of these racial funding disparities and their impacts. Likewise, examining funding rate disparities by research discipline also adds crucial context.

NSF divides its research and education portfolio into seven grant-making directorates: Education and Human Resources (EHR); Social, Behavioral, and Economic Sciences (SBE); Biological Sciences (BIO); Geosciences (GEO); Computer and Information Science and Engineering (CISE); Engineering (ENG); and Mathematical and Physical Sciences (MPS).

All award and decline decisions are issued through program offices and divisions specializing in distinct subfields within each directorate. Aside from differences in scientific purview, each directorate also handles varying numbers of proposals and funds them at different rates depending on their budget. As a result, overall funding rates differ between directorates, with some more competitive than others (e.g., 14.

8% in EHR versus 25. 5% in GEO for Research, 2012–2016). Despite these differences, from 2012 to 2016 (years with available data), after normalizing by overall directorate funding rates, we find that all directorates exhibited racial funding rate disparities and stratification patterns, albeit to varying degrees.

Most patterns for overall funding rates were also observed at the directorate level. In every directorate, proposals by white PIs were consistently funded above overall directorate funding rates, regardless of type ( Figure 4A ), and for all non-white groups, relative funding rates for Research proposals were also below those for Non-Research, with rare exceptions (e.g., GEO for Black/AA PIs).

The proportion of Research awards to Black/AA PIs was also consistently below overall directorate proportions across all directorates ( Figure 4C ). Within each directorate, relative funding rates for proposals by white and Asian PIs exhibited less year-to-year variability, owing to larger submission numbers.

For other groups with fewer proposals, funding rates were more volatile, and in the case of AI/AN and NH/PIs, data were often missing (in most merit review reports, proposal or award sums fewer than 10 were omitted to protect the identities of individual investigators).

Figure 4 with 22 supplements see all From 2012 to 2016, all disciplinary directorates exhibited racial disparities in funding rates and racial stratification in awarded activities. ( A ) Relative funding rates by directorate for Research (top) versus Non-Research (bottom) proposals by PI race and ethnicity. Gray circles mark relative funding rates for each available year; colored rectangles represent the multi-year average.

To aid visual comparison, the multi-year average relative funding rate for Research proposals is superimposed on the Non-Research panel as a dotted rectangle. For Research proposals, data are available for at most 5 years (2012–2016); for Non-Research proposals, data are available for at most 4 years (2013–2016).

( B ) Multi-year average annualized award surplus or deficit per directorate by PI race and ethnicity, for Research (top) and Non-Research (bottom). The upper-left number in each sub-panel is the multi-year average annualized award surplus or deficit for each group for all seven directorates, excluding awards made by the Office of the Director.

For AI/AN and NH/PI PIs, only data for Research awards in 2012 are shown; no directorate data for Non-Research awards are available. ( C ) Proportion of awards for Research by directorate and PI race and ethnicity, compared to overall directorate proportions (black horizontal lines), 2013–2016. White text denotes average annual number of Research awards per directorate to each group.

( D ) Percentage of all proposals submitted to each directorate by white (red), Black/AA (blue), Asian (green), and Hispanic or Latino PIs (yellow) versus the multi-year average relative funding rate for all proposals by each group, 2013–2016. Source data: Data S4 in the accompanying data repository ( https://doi. org/10.

5061/dryad. 2fqz612rt ). Between directorates, the magnitude of disparities and the group with the lowest funding rate varied.

For Research proposals with available data, Black/AA PIs had the lowest multi-year average funding rate in CISE, EHR, ENG, SBE, and MPS, whereas Asian PIs had the lowest in BIO and GEO. Considering disparities by each year, while Research proposals by Black/AA PIs were consistently the lowest funded in CISE for every year between 2012 and 2016, the group with the lowest funding rate occasionally changed in other directorates.

In comparing the magnitude of disparities for Research proposals across directorates, white PIs experienced the largest funding rate advantage over Asian PIs in BIO (1. 5-fold, multi-year average) and the largest advantage over Black/AA PIs in SBE (1. 7-fold, multi-year average; Data S14).

The impact of these disparities in terms of the award surpluses and deficits broadly scales with directorate size, or more specifically, by the number of proposals managed by each directorate. For example, while the +9.

7% average relative Research funding rate for white PIs in MPS was not the highest of all directorates, because 21% of all Research proposals by white PIs were submitted to MPS in this period, this elevated funding rate accounted for 26% of the total Research award surplus to white PIs (117 of 441, multi-year annualized average; Figure 4B ). However, especially high or low relative funding rates have amplifying effects.

For example, although BIO received only 6% of all Research proposals by Asian PIs, because its relative Research funding rate for Asian PIs was very low (–27. 8%, multi-year average), the lowest of all directorates during this period, BIO contributed 11% of the total Research award deficit to Asian PIs (34 of 306, multi-year annualized average).

These directorate funding data also reveal a paradoxical trend: relative funding rates for proposals by Black/AA PIs are lower for directorates with proportionally more proposals from Black/AA PIs ( Figure 4D ), with the exception of EHR. Although 2. 4% and 2.

5% of all proposals to SBE and ENG were submitted by Black/AA PIs, the multi-year average relative funding rates for proposals by Black/AA PIs in SBE and ENG were the lowest of all directorates, at –26. 5% and –19. 6%, respectively (Data S4).

The same trend is observed for Research proposals ( Figure 4—figure supplement 1 ). To guide funding decisions, NSF program officers within each directorate oversee the vast majority of proposals through a 6-month-long external peer review process, wherein outside experts with field-specific expertise provide feedback on the merits of a proposed project.

Through individual written input and/or panel deliberations, external reviewers are instructed to assess a proposal’s potential to advance knowledge (intellectual merit), its potential to benefit society (broader impact), and the qualifications of the PI, collaborators, and institution ( National Science Foundation, 2021 ).

In addition to narrative comments, external reviewers must also give an overall rating on a scale from ‘Poor’ (numerically 1, “proposal has serious deficiencies”) to ‘Excellent’ (numerically 5, “outstanding proposal in all respects”). A minimum of three pieces of external input are required for complete evaluation.

While self-reported demographic data are not visible to the reviewers, PI race or ethnicity may be inferred from proposal content or personal knowledge. Data on average review scores of externally reviewed Research proposals show that proposals by white PIs received higher scores than proposals by all other non-white groups, with scores negatively skewed, asymmetrically distributed towards higher ratings ( Figure 5A ).

In 2015, the average of all average review scores for white PI proposals was 3. 46 (median 3. 50), compared to 2.

98 for Black/AA (median 3. 00), 3. 10 for NH/PI (median 3.

00; score distribution unavailable), 3. 11 for AI/AN (median 3. 33), and 3.

23 for Asian PI proposals (median 3. 25). Similar differences in average review scores are observed in 2016, the only other year with available data.

In 2015 and 2016, the distributions of average external review scores of externally reviewed Research proposals were systematically offset and skewed based on PI race. ( A ) White (red dashed), Black/AA (blue), Asian (green), and AI/AN (purple) PIs, for 2015 (left column) and 2016 (right column). Proposals are rated on a scale from 1 (Poor) to 5 (Excellent).

The grand average of all average review scores for proposals by white PIs (red-outlined arrows) is higher than the grand averages of review scores for proposals by Black/AA, Asian, and AI/AN PIs (solid-colored arrows). ( B ) Funding rates of externally reviewed Research proposals by average review score and PI race. For context, the funding rate of all Research proposals by group is listed from highest to lowest in the top left corner.

Data on review scores and funding rates by average score are only available for 2015 and 2016. Source data: Data S5 in the accompanying data repository ( https://doi. org/10.

5061/dryad. 2fqz612rt ). Accompanying information on the success rates of proposals based on average review score highlights the impact of programmatic decision-making.

In 2016, although average scores for Research proposals by Black/AA PIs were lower, the relative funding rate for Research proposals by Asian PIs was worse, –20. 9% for Asian PIs compared to –17. 3% for Black/AA PIs ( Figure 3C ).

This counterintuitive result may be attributed to differences in success rates for proposals with comparable scores. Although proposals with higher scores are more likely to be awarded, the success rates for Research proposals by Black/AA PIs were generally higher than those for white and Asian PIs with the same score ( Figure 5B ). For Asian PIs, success rates for Research proposals are not as high, closer to the rates for white PIs.

These decisions to fund proposals outside of their rank order by score reflect NSF’s discretion to consider scores alongside other factors when making funding decisions, such as reviewer comments, panel discussion summaries, and a need to balance a diverse research portfolio in line with the agency’s statutory mission and national interest ( National Science Foundation, 2021 ).

Although proposals by white PIs generally experience lower success rates by score compared to most other groups, the large absolute number of proposals by white PIs combined with their above average scores still resulted in relative Research funding rates of +8. 2% and +12. 0% for white PIs in 2015 and 2016.

These racial funding disparities raise many questions about their underlying causes and mechanisms, but limitations of current publicly available data reported by NSF restrict such inquiries from being robustly investigated. Since 2003, NSF has used the racial and ethnic categories and definitions set by the Office of Management and Budget in 1997, following government-wide standards for federal data collection.

However, racial and ethnic categories are understood to be social constructs with no biological basis, and as such, are complex and highly mutable over time, subject to changes in social perceptions of race and self-identification ( Clair and Denis, 2015 ).

Furthermore, because our data are limited to those published in annual merit review reports, which change in content and organization each year, disaggregated information on funding outcomes by directorate and award type are only available for short time intervals, limiting our ability to fully characterize long-term trends.

Lastly, modifications to the way race and ethnicity information are tabulated in merit review reports also impact data consistency (see Methods; Figure 6—figure supplements 1 – 2 ). The lack of publicly available data also precludes multivariate and intersectional examinations of NSF racial funding disparities alongside other factors like gender, career stage, and institution type.

Although NSF data on funding rates for PIs by these aforementioned characteristics exist, this information is tabulated separately from data by PI race and ethnicity (e.g., Figure 6—figure supplements 3 – 5 ).

Due to lack of data access, we are also unable to investigate the influence of other factors that likely affect funding rates, like educational background and training, prior scholarly productivity as publications, previous funding success, mentoring networks, and institutional knowledge and support. Many of these factors have been previously shown to add crucial context to the racial funding disparities at NIH (e.g., Ginther et al.

, 2011 ; Ginther et al. , 2018 ). Another consideration is that data on funding outcomes by PI race and ethnicity are reported only as total numbers of proposals and awards, and do not include information on the number of unique applying PIs in each group.

In any given year, a PI can submit multiple proposals and likewise receive multiple awards, making per-PI race-based differences in submission or award rates indiscernible from the available data.

However, this information is reported in aggregate for Research grants in three-year windows ( Figure 6—figure supplement 6 ): for example, in 2017–2019, approximately 52,600 unique PIs submitted a total of 114,655 Research proposals, and of these PIs, approximately 39. 4%, or 20,700, received at least one award.

For every three-year window since 1995, 34–44% of all PIs who applied for at least one Research grant received at least one award. Of these funded PIs, approximately 13–16% received two awards, and 4–5% at least three.

We also observe an emerging trend in the non-reporting of demographic information: from 1999 to 2020, the proportion of proposals submitted by PIs who provided information on their race decreased from 96% to 66% ( Figure 6 ). This trend is accelerating, with a 10% drop in response rate between 2019 and 2020, the largest year-to-year decrease observed in available data.

This pattern coincides with similar decreasing trends in the response rate for ethnicity and gender ( Figure 6—figure supplement 7 ). The cause of this phenomenon and its prevalence elsewhere is unclear, as it has not been widely reported.

Regardless, these trends are concerning, as further decreases in the proportion of respondents will undermine the statistical effectiveness of reported information, impeding future efforts to track disparities. Figure 6 with 7 supplements see all The decline in the proportion of proposals by PIs who identified their race has accelerated in recent years.

Source data: Data for 1999–2019 are collated in Data S1 in the accompanying data repository ( https://doi. org/10. 5061/dryad.

2fqz612rt ); data for the year 2020 are available from the 2020 NSF report on the Merit Review Process ( National Science Board, 2021a ). Our analysis shows that for at least two decades, there has been a consistent disparity in funding rate between proposals by white PIs and those by most other racial groups. The relative funding rate for proposals by white PIs has also been increasing with time.

We further show that disparities are even greater for Research awards, a result obscured within overall statistics by higher Non-Research funding rates. Differences in the allocation of awards for Research versus Non-Research activities by racial group reveal a stratification of funded activities by race. These patterns are also observed within each directorate.

Identifying the underlying causes and mechanisms for these disparities requires further study, but information on average external review scores from two recent years sheds light on processes that influence outcomes. The racial funding disparities at NSF are comparable in magnitude, persistence, and aspect to those found in other funding bodies, despite differences in internal review processes and discipline-specific norms.

In some cases, these patterns have notable similarities. For example, the 1. 7- to 1.

8-fold advantage for NSF Research proposals by white PIs over those by Black/AA PIs in 2013 and 2014 was likewise observed for NIH “R01”-type research proposals in 2000–2006 ( Ginther et al. , 2011 ), 2011–2015 ( Hoppe et al. , 2019 ), and 2014–2016 ( Erosheva et al.

, 2020 ), with the same 1. 7- to 1. 8-fold magnitude.

Our finding that racial disparities persist at the directorate level is consistent with an NIH study showing that Black/AA PIs experience both overall and within-topic funding rate disadvantages compared to white PIs ( Hoppe et al. , 2019 ). At the National Aeronautics and Space Administration (NASA) from 2014 to 2018, proposals by white PIs were funded at rates 1.

5 times higher than those by underrepresented racial and ethnic minorities (defined by NASA as AI/AN, Black/AA, NH/PI, multiracial, and Hispanic or Latino PIs), and were considered “consistently over-selected” at “above reasonable expectation” in 2015, 2016, and 2018 ( National Academies of Sciences, Engineering, and Medicine, 2022 ).

Outside the US, widening gaps in funding rates and award amounts between white and ethnic minority PIs have been documented at the Natural Environment Research Council and the Medical Research Council, the UK counterparts to NSF and NIH ( UK Research and Innovation, 2020 ). Similarly, the UK-based global research philanthropy Wellcome Trust reported a 1.

9-fold disparity in funding rate between white and Black applicants from 2016 to 2020 ( Wellcome Trust, 2021 ). This finding