It is fair to say that salary is an obligatory factor in every study that explores whether there are differences across gender in academic careers. Faculty salaries have been the subject of numerous university salary equity investigations, occasional lawsuits, and broader national studies. (See, for example, selected works by Barbezat, Becker, Bellas, Benjamin, Farber, Ferber, Ginther, Johnson, Perna, and Toutkoushian in the bibliography.) In general, studies suggest that women’s salaries tend to lag behind men’s. This, for example, is the conclusion that one would draw from the salary data that were collected by the American Association of University Professors (see Chapter 1). Data collected by the American Association for the Advancement of Science’s salary and job survey (Holden, 2001, 2004) also support that claim. The gap, however, appears to be shrinking and our data, discussed below, confirm this tendency. Furthermore, while at first glance salary would appear to be a well-defined quantity that can be easily compared across gender, many factors appear to affect salaries in a complex way. Therefore, it is important to account for the potentially confounding effects of factors such as discipline, rank, productivity and others before attributing possible salary discrepancies to the effect of gender.
Here we examine salary information collected as part of our survey, as well as the salary data included in the NSOPF:04.
The faculty survey asked respondents to report their base salaries. We consider only the 1,404 full-time faculty who responded to the survey and who were assistant, associate, and full professors. There were 1,179 faculty for whom the salary information was not missing.[79] Appendix 4-21 shows the number of missing salary observations in each discipline and by gender. As is clear from the table, the proportion of faculty who did not respond to this question is similar across gender and across disciplines. The four observations correspond to two men and two women, and all exceeded $600,000 for a 9-month salary. The next-highest salaries reported were all below $250,000 for 9 months. One of the four outliers that were removed corresponded to a reported salary of almost $1.8 million, which is clearly unrealistic. About 20 percent of all respondents reported salaries below $100 for 9 months of work. Since these are likely to be values reported as thousands, we decided to multiply those reported salaries by 1,000 rather than lose the information. Two other salaries were removed from consideration and corresponded to two faculty members who, even after rescaling, ended up with 9-month salaries below $ 10,000 (the next lowest salary was $45,000). The wisdom of deleting the four highest salaries from the data set might be debatable, but from a purely statistical viewpoint is fully justified; the next-highest 9-month salary in the sample was $212,272, so the salaries at the high end of the distribution were clear outliers. Similarly, at the lower end, there was a clear gap between salaries below $10,000 and the next lowest at $45,000. Thus, for our initial salary analyses, we considered 1,173 faculty out of the 1,179 who responded.
Appendix 4-9 shows the mean salary by discipline, rank, and gender. It also shows the number of observations in each category. Statistics were computed using salaries standardized to a 9-month basis. Nor surprisingly, salary increases with seniority in all disciplines and both genders. Men appear to have a higher mean salary than women in almost all disciplines, but only among full professors. The difference between men and women seems to vanish for associate and assistant professors, and in some disciplines (e. g., electrical engineering and physics), female associate professors appear to receive a higher mean salary than their male colleagues. At the assistant professor level, the differences in mean salary are negligible and favor men or women, depending on the discipline. One interesting finding is that the highest salary among assistant professors is paid to a woman in every discipline, while the lowest salary is paid to women in only half the disciplines (mathematics, physics, and civil engineering).
In trying to understand the major predictors of salary, we first fitted a simple model that did not take into account potentially important factors, such as productivity. In this simple model, explanatory variables were gender, rank, academic age, discipline, and all the two-way interactions with gender. There were 1,169 observations with complete covariate information, and the model fitted the data well: R2 = 0.54.
There were significant gender differences in salary in this model (p = 0.009), in which we controlled for several variables likely to differ between men and women (e. g., rank and discipline). However, the effect of gender cannot be interpreted in isolation of other factors because the interaction between gender and rank was also statistically significant (p = 0.004). Thus, we can only investigate whether salaries for men and women are similar within rank. We find that among full professors, men earn significantly more than women (p < 0.05). On average, male full professors earn about 8 percent more than female full professors. There are no significant differences in salaries for men and women among associate or assistant professors. Discipline and rank were also significant predictors of salary. Given these results, it seems likely that some of the gender differences in faculty salaries that are reported in other studies, in which rank and discipline are not controlled, are due more to the confounding factors rather than solely to any gender difference in salary.
We also looked at the salary data in the NSOPF:04. Because income is reported as a mean and there are small sample sizes in some disciplines, it was not possible to break out the analysis by field. In the NSOPF :04 data for full-time faculty who had instructional duties for credit and faculty status at an Research I institutions in engineering, biological sciences, physical sciences, mathematics, and computer sciences, we found that men received marginally larger salaries than women at the full professor level. However, there was no significant difference between the salaries of male and female faculty at the associate professor or assistant professor level. This assessment of gender differences in base salaries did not control for disciplinary differences or academic age, both of which are likely to have gender differences in them.
We were also interested in whether several additional variables predicted salary. To explore this, we again fitted a linear model to the log of base salary (9-month base), but now extended the list of explanatory variables in the model to include gender, discipline, faculty rank, type of institution (public or private), prestige of the institution, grant funding, publications (refereed journals and conference proceedings), academic age (defined as the time elapsed between award of the Ph. D. and December 2004), and all two-way interactions with gender. The model was fitted to the 753 observations with complete information for all covariates and resulted in an R2 equal to 0.64. Gender, discipline, rank, institution type, and prestige were considered classification variables; the remaining were included as continuous variables. A random effect for institution was included, but the institution variance component was negligibly small.
Gender was not significantly associated with salary once other potential confounders were taken into account. Significant associations with salary were found for discipline (p < 0.0001), rank (p < 0.0001), prestige of the institution (p < 0.0001), type of institution (p < 0.0001), and grant funding (p = 0.002). The interaction between rank and gender was again significant (p = 0.02). Academic age and the interaction between grant funding and gender were approaching significance (p = 0.07 and p = 0.09, respectively).
As would be expected, full professors reported larger salaries than associate professors, who reported larger salaries than assistant professors. The more time that had elapsed since a faculty member received a Ph. D., the higher the salary, regardless of rank. The highest prestige institutions across all disciplines pay higher salaries than medium-prestige institutions, which in turn pay higher salaries than the lowest prestige institutions. Private institutions pay higher salaries than public institutions.
There was a positive association between grants and salary. Everything else being equal, a faculty member who increases his or her funding by a factor of 2.7 in a year would be linked with a 0.6 percent increase in salary. For a $75,000 annual salary, this would amount to an increase of $450, and so while this relationship is statistically significant, it has little practical significance. While the association between the number of publications and salary was not statistically significant, it was at least positive. Once again, please note that the discussion here somewhat overshoots the capabilities of our surveys. Because our surveys were cross-sectional, a conclusion stating that a faculty member who, over time, increases funding will also see an increase in her salary implies a longitudinal effect that our data cannot capture. Thus, these results, while plausible, must be cautiously interpreted.
Although women and men at a given rank appear to be compensated at similar levels (this does not apply for full professors), women may be at a disadvantage if they are less likely to be promoted to higher ranks. This topic is addressed in Chapter 5, where we found no evidence of differences among men and women in terms of promotion to higher ranks in our sample of full-time faculty at Research I institutions.