As we mentioned in a prior post, this time we’re shifting our focus to Doctoral Dissertation Improvement Grants, DDIGs.
DEB has supported DDIGs since 1970 but our electronic records only go back to 1979. Since there appear to be multiple differences in how DDIGs were prepared, submitted, and/or transferred to electronic records prior to 1983 and 2014 grants are still in the awarding process, we limit our analyses to the 1983 to 2013 period[i].
Basic DDIG Data:
The DDIG program has grown fairly steadily since the early 1980s with the steepest growth over the last decade. The success for these proposals has remained about 30% with peaks in the late 1980s and early 2000s. Interestingly, the rise and fall of the success rate data are consistent with responses to major periods of national concern over science student shortages[ii]. These fluctuations appear to even out in the long run with most recent 3 years’ success rates (28%, 32%, and 30%) comparing well with the long-term success rate of 31%.
The best years on record by success rate were 1990 (38%) and 2002 (37%). The worst years were 1995 and 2010 (both 25%).
We hope you look at the above figure and take note of that funding rate. It’s not doing anything like what we see in our core programs and elsewhere in research funding. As the number of submissions has grown, the success rate has stayed essentially constant. This is because we really like DDIGs; they are an important investment in terms of science and in terms of capacity building. And, as you often like to tell us, DDIGs have an amazing impact to cost ratio. We can do this because DDIGs are relatively cheap; even though the absolute dollar amounts for DDIG projects has grown, they are still small compared to regular research grants and we can provide the necessary support without causing a big pinch elsewhere in our portfolio.
The DDIG records identified 6226 student Co-PIs with unique PI identification numbers submitting between 1983 and 2013, though some early records were missing this information. Of these applicants, 2439 eventually received support for a DDIG project. If we consider the individual student applicants instead of proposal jackets, the DDIG program has an overall success rate of 39%[iii].
A large majority, 85%, of applicants report their gender as part of their PI profile. Viewing the outcomes of the DDIG program through this filter, we see that female students have been funded for DDIGs in proportion to their applications, though male students make up a larger portion of the sum total of both proposals and awards since 1983.
The proportion of female students in the annual pool of Co-PI applicants and awardees has roughly doubled since 1983. Since the mid-2000s female students have held the narrow majority (for students reporting gender) in both proposals and awards.
Even though the proportion of female students on DDIGs is growing, it is not as a result of fewer male students receiving DDIG awards. The number of annual submissions (not shown) and awards (shown below) for male students lacks any directional trend for the 1983-2013 period. Instead, the growth in the DEB DDIG program has been driven almost entirely by increases in the number female student awardees over the past 30 years.
Longer Term Outcomes:
Do NSF proposal data show any long-term benefits to DDIG students?
As Co-PIs on DDIG proposals, every student applicant creates a unique profile that they carry with them and update over the course of all subsequent proposal applications to NSF. In order to look at what happens after a DDIG proposal, we pulled records for subsequent submissions associated with the student’s profile. We looked at two cases: later research proposals where the student was listed as PI, and later DDIG proposals where the student had become an advisor (2nd Generation DDIG PIs). We could then look for differences in the proportions of students returning as PIs.
For all DDIG student applicants over the 1983-2013 period, just over 25% have appeared on subsequent research proposals and the return rate as subsequent DDIG advisors is near 8%.
Male students were 50% more likely than female students to return as research PIs and twice as likely to return as future DDIG advisors. However, this ignores the changing demographics of the program which means the “average” female applicant has had less time to accumulate additional PI experience than the “average” male applicant.
To eliminate this bias, we made an age-controlled comparison looking at applicants from the first decade of the data, 1983 through 1992. Looking only at these older applicants we see both higher return rates and much smaller differences between males and females.
Those who received DDIG funding are more likely to come back to NSF for funding in the future as either a research proposal PI or as the advisor to a subsequent DDIG applicant, compared to DDIG applicants who did not receive funding over the 1983 to 2013 period.
In this case, we’ve again used the 1983 through 1992 age-controlled comparison group.
In all cases, DDIG support is associated with higher future return than their unfunded peers. Male students have been more likely than females to return seeking support regardless of funding outcome. The association between DDIG support as a graduate student and returning as a research PI appears similar for males and females (Funded Student Research PI return rate ~1.6x Unfunded Student Research PI Return Rate). Funded male students show a relatively larger increase over unfunded peers than do female students in the case of returning as a DDIG advisor.
In our next DDIG post, we will examine how DDIG support relates to some other measures for career trajectories.
[i] Our previous Numbers posts have generally stuck to looking at recent trends. This is because coding practices and recording methods have changed over time and it’s difficult to verify and sort older records and wade through the sheer volume of materials for quality control. This is especially true of anything from before the transition to full electronic submission in 2000 since the database entries would have to be compared to physical records in an offsite warehouse (think Raiders of the Lost Ark).
[ii] Specifically, Teitelbaum’s “Round 3” ending in the early 1990s and “Round 4” and “Round 5” overlapping but ending in the early 2000s.
[iii] Even accounting for missing data, # applicants < # proposals due to resubmission in subsequent years.