In the sciences and engineering, grant activity is an important demonstrator of research ability. There are a number of approaches to comparing male and female faculty as grant recipients. Basic measures are whether faculty have received any grants and the total dollar value of any grants received. More in-depth measures include whether the grantee is a principal investigator (PI) or co-principal investigator (Co-PI), the source of grants, and the type of grants.
NSOPF:04 asked respondents whether any of their scholarly activity was funded. One hundred percent of both male and female full-time, professoriate faculty in S&E at Research I institutions indicated that it was. Our faculty survey found a lower percentage of faculty who responded that they were receiving sponsored research grants. It asked respondents, “What is the total dollar amount of the research grants on which you served as principal investigator or co-principal investigator during the 2004-2005 academic year? Include only direct costs for academic year 2004-2005.” There were 213 faculty out of a total of 1,404 fulltime faculty who did not provide an answer to the question. Of the 1,191 respondents, 163 (14 percent) answered zero, which means 86 percent of respondents (and 73 percent of all full-time faculty) reported some grant funding. This lower percentage found in our faculty survey compared to the NSOPF:04 is probably due to the fact that we asked faculty to limit their response to grants on which they served as a PI or Co-PI. Further, not all faculty may have interpreted the question in the same way; it may be that some among those reporting no funding had research funding during 2004 to 2005 but did not receive any new funding. A similar proportion of men and women (16 percent and 14 percent, respectively) had missing grant information. Thus, we base the remainder of the discussion on those faculty members who reported either no grant funding or some grant funding in the survey.
Women were more likely than men to report that they had grant income (88.6 percent compared to 83.8 percent). The University of Pennsylvania gender equity report found that women and men were equally likely to obtain grant support. As shown in Figure 4-8, disaggregated by field, women and men were equally likely to have at least one research grant on which they served as a PI or Co-PI, except in civil engineering (women, 99 percent; men, 88 percent) and mathematics (women, 77 percent; men, 62 percent), in which the differences were significant.
However, there may be differences when grants are examined in more depth. Although the University of Pennsylvania gender equity report found that men and women were equally likely to obtain grant support, it also found that men were more likely to be PIs, which suggests that an important focus for research would be which faculty have been PIs and which have been Co-PIs. Second, men and women may not be receiving the same types of grants. For example, while women’s participation in National Institutes of Health grants is growing, the percentage is still quite small, and in some categories, the size of awards of the same type are smaller for women than for men (OER, 2005). However, a recent study by RAND (Hosek et al., 2005) found no gender differences in the amount of funding requested or awarded during the period 2001 to 2003 at National Science Foundation (NSF) and the U. S. Department of Agriculture (USDA). Interestingly, the study also found “gender differences in the fraction of first-year applicants who submit another proposal in the following two years” (p. xii). Women were less likely to reapply.
The next question in the faculty survey focused on the size of the grants received by the faculty member. About 6 percent of respondents answered that they had received $1 million or more in grant funding. The median response was $160,000.[77]
To explore the association between individual and institution-level factors and the success with which faculty raise research funding, we proceeded in two steps. We first modeled the binary outcome “grants/no grants” using logistic regression as a function of the following covariates: (1) gender, (2) disciplinary area (e. g., biology), (3) faculty rank (assistant, associate, or full), (4) type of institution (e. g., public versus private), (5) prestige of the institution, (6) number of publications,
|
Biology Chemistry Civil Electrical Mathematics Physics Engineering Engineering |
□ Men □ Women
FIGURE 4-8 Percentage of faculty reporting having at least one research grant on which they served as a PI or Co-PI by gender and field.
SOURCE: Faculty Survey carried out by the Committee on Gender Differences.
(7) percent of time spent on research, (8) whether the faculty member reported that he or she had a mentor or not, and (9) all two-way interactions between the above covariates (see Appendix 4-20a and 4-20b for the analysis). This provided an estimate of the chance that a faculty member would or would not receive a research grant, regardless of the size. It is important to mention that whether a faculty member was or was not awarded a grant is of interest in itself because in some disciplines, receiving funding from a competitive agency is at least as important as the actual amount of funding received.
In a second modeling step, we estimated the amount of funding received conditional on having at least some research funding. The dependent variable was not the amount of grant funding, but instead the logarithm of the amount of grant funding to provide a dependent variable with a less skewed distribution, which can be useful in such models. There were 799 faculty (out of 1,191 who responded to the question about grants) with complete information for all model covariates. Of these, 697 (87 percent) reported receiving some grant funding during the period of interest.
We cannot conclude whether gender is associated with the probability of having a grant because the interaction between gender and discipline and between gender and rank were both statistically significant (p < 0.05). Women were significantly more likely (p < 0.05) to report having some grant funding in mathematics and in civil engineering, but the differences between the two genders were not significant in the other disciplines. Female full professors were significantly more likely to report some grant funding than their colleagues at the associate and assistant professor ranks, but for men, the gender rank interaction term was significant (see Appendix 4-20a and 4-20b). Overall discipline was significantly associated with the probability of having a grant, with faculty in electrical engineering being less likely (p < 0.05) to have a grant than faculty in the other disciplines. Faculty in civil engineering were significantly more likely (p < 0.001) than faculty in other disciplines to report some grant funding. The effect of discipline, however, is impossible to isolate, since the interaction between discipline and gender is highly significant, even after accounting for confounders such as rank, type of institution, and others. Assistant professors were less likely than associate professors (p = 0.007) and full professors (p < 0.0001) to have a grant, but there was no significant difference between full and associate professors. Again, discussing the effect of rank independently of gender is not reasonable, given the significant interaction between the two factors. Faculty at private institutions were equally likely to have a grant than those at public institutions (p > 0.05), and faculty at institutions of lower prestige were only marginally less likely (p = 0.06) to have a grant than faculty at institutions of either medium or higher prestige (which did not differ from one another). The number of publications a faculty member had was not associated (p = 0.9) with the probability of having a grant. Faculty who spent a greater percentage of their time on research were more likely to have a grant (p < 0.01). Finally, a faculty member who had a mentor appeared at first glance to be less likely to have a grant than a faculty member who did not have a mentor (p < 0.0001). However, this effect is difficult to interpret because the beneficial effect of the mentor depended on the gender of the faculty being mentored (p < 0.03). In fact, and contrary to what we might have anticipated, survey results suggest that the effect of having a mentor is not statistically significant among men. However, among women, a strong association between having a mentor and having grant funding was demonstrated.
Regarding the data on the interplay of gender and the availability of a mentor, we find in Table 4-2 that female assistant professors who do not have a mentor have a substantially lower probability of having a grant than female assistant professors who do have a mentor.[78] For male assistant professors, the presence of a mentor seems to make little difference. For associate professors, the presence of a mentor is associated with an increase in the probability of receiving a grant for both men and women, but the effect is much less pronounced than for female assistant professors.
We now consider the size of the grant and model the amount of funding as
|
TABLE 4-2 Percentage of People Who Received Grant Funding by Gender, Rank of Faculty, and Mentor Status
SOURCE: Survey of faculty carried out by the Committee on Gender Differences in Careers of Science, Engineering, and Mathematics Faculty. |
a function of the same covariates. We used a log transformation on the response variable to better meet the normality assumption in the linear model. Because the log transformation collapses at zero, we added a negligibly small amount ($10) to the funding reports of zero. Figure 4-9 shows the distribution of funding amounts for all disciplines (except mathematics) in the log scale. Note that the distribution has a point mass at 2.3, which corresponds to the log of 10—the small amount added to grants of zero.
To explore the association between gender and other covariates on size of the grant, we considered all observations but fitted a Tobit regression model where the truncation bound was set to 3.0 (since the log of 10 is 2.30). That is, we obtained estimates of regression coefficients in the model that are unbiased and consistent once we account for the truncation. Of the observations, 485 exceeded the lower truncation bound and 221 did not.
There was no difference in the amount of grant funding received by male and female faculty after accounting for possible confounders of discipline, rank, type of institution, prestige of the institution, and research productivity (as measured by the number of publications).
Faculty in mathematics received grants of significantly smaller size (p < 0.001) than faculty in all other disciplines. Mathematics aside, the differences among all other disciplines were not statistically significant, with the exception of biology. Full professors had significantly larger grants than associate professors (p < 0.0001), who in turn had significantly larger grants than assistant professors (p < 0.003). Faculty at universities of highest prestige had significantly more grant funding (p-value = 0.002) than faculty at institutions of lower prestige, but the difference between the size of the grants received by faculty at the highest and the medium-prestige institutions was not significantly different (p = 0.12). Faculty at institutions of medium prestige had, in turn, marginally more funding than faculty at institutions of lower prestige (p = 0.07). These results are supported by the following average grant sizes: The average size of a grant obtained by faculty in 2004-2005 was $336,257, $352,639 and $463,231, respectively, at institutions of lowest, medium, and highest prestige. These values must be interpreted with caution. Making sweeping inferences
|
|
|
|
 |
|
|
|
 |
|
logrant MIDPOINT
(Natural) Logarithm of Grant Amount in Dollars
FIGURE 4-9 Histogram of the logarithm of grant amount (observed distribution of grant values in the log scale).
about funding levels across institutions, ranks, or even gender is unwarranted, given that we are considering only 1 year of funding data. Neither the number of publications nor the percent of time spent on research activities were associated with the size of grants obtained by faculty (p = 0.3 and p = 0.7, respectively); faculty with a mentor had less funding than faculty without a mentor (p = 0.01), which may reflect the fact that mentors are more prevalent among younger faculty who in turn tend to receive the smaller grants.
It is also of interest to investigate the association between gender and covariates conditional on funding. That is, if we were to consider only those faculty members who reported receiving some funding during 2004 to 2005, would results differ from those obtained when analyzing the entire set of outcomes? We anticipated that gender would not be associated with the amount of funding received by a faculty member even in the conditional analysis, given that gender was not found to be a predictor of the probability of receiving a grant or of the amount of grant funding unconditionally. A multivariate normal regression model fitted to the log-transformed positive grant values leads to approximately the same conclusions as the analysis that considers both zero and positive values together. Gender was still not associated with the amount of funding received, given that at least some funding was received. Full professors received significantly more funding than associates, who in turn received more funding than assistant professors. As before, the prestige of the university was positively associated with the amount of funding; the higher the prestige, the higher the average size of grants, everything else being equal. As before, we found that discipline was significantly associated with grant size, but essentially all of this effect is due to the fact that faculty in mathematics received significantly smaller grants than those in the other five disciplines.