Ph.D.-granting departments in the U.S. claim to have increased the number of female faculty, to be graduating more Ph.D. students, and to have again increased their undergraduate enrollment, especially the enrollment of female students. This Viewpoint is derived from a report that used Taulbee Survey data from 19992009 to provide insight into these claims and other questions regarding gender demographics. In particular, we take a close look at whether the trends are different for private and public institutions, for large and small departments, and for departments of different ranks.
The data indicates that overall both the number and proportion of female tenured and tenure-track faculty as well as females receiving a Ph.D. have increased; however, these increases are not uniform across types of departments. In many highly ranked departments female full professors outnumber female faculty of other ranks and increases have been small. Departments that have significantly increased the number of female faculty tend to have done so by hiring female assistant professors. Although the overall Ph.D. production has doubled from 20022008, the proportion of Ph.D.'s awarded to females is lower in higher ranked departments. In contrast, higher ranked private institutions have a higher proportion of female undergraduates. This Viewpoint highlights some of these results and trends generated from Taulbee data; the full report is available at http://www.cs.purdue.edu/homes/seh/Taulbee99-09_Full.pdf.
The Taulbee Survey, conducted annually by the Computing Research Association (CRA), provides information on Ph.D., M.S., and B.S. enrollment and production in computer science, computer engineering, and information schools in Ph.D.-granting departments in North America. Results of the Taulbee Survey since 1992 are available at http://www.cra.org/resources/taulbee/. The Taulbee Survey uses the 1995 National Research Council (NRC) rankings to form four groups: The 36 top-ranked departments are partitioned by their rank into three groups, 112, 1324, and 2536. The remaining departments belonging to the CRA, some ranked and some unranked by the NRC, form one group (referred to as 37+).
Due to the increase in the overall number and proportion, the number of females receiving a Ph.D. has more than doubled since 2002.
To refine the 127 departments in Taulbee's group "37+" and to provide insight into trends of public versus private institutions, our study uses a different partitioning of the 163 U.S. CS and CSE Ph.D.-granting departments. Groups had to be large enough to ensure no department could be identified and we wanted to retain some ability to compare our groupings against Taulbee Survey results. The underlying eight groups are formed as follows.
The accompanying table shows the number of departments in each of the eight groups and the average response rate to the Taulbee Survey. A listing of the departments in each group is available in the full report.
Since 1999, the number of faculty in CS and CSE departments in the U.S. has increased. For the 163 departments considered, the average number of faculty increased from 22 in 1999 to 28 in 2009. The number of female tenure-track and tenured faculty has also increased. In proportion, it increased from 11% to 14.8%. Looking at the eight groups, the average numbers for female faculty range from a minimum of 1.13 (for departments in group Public 37+ Medium in 2001) to a maximum of 7.57 (for Public 136 2nd in 2009). The next two figures provide insight into the increases for female faculty for the eight groups, one for proportions and one on average numbers for female faculty in the three academic ranks.
Figure 1 shows that each group has seen an increase in proportion in female male faculty. Each panel shows a line representing the proportion of female faculty and bars indicating the difference from the overall average for the year. The average proportion of female faculty of 163 institutions forms the "baseline" of each year. The eight panels show that increases are not uniformly distributed across groups.
In order to better understand increases in proportion and increases in numbers, we examined the average number of female faculty in each rank. Figure 2 shows the average number of female faculty in each group, partitioning the total number of female faculty into assistant, associate, and full professors. As Figure 1 shows, group Public 136 2nd has seen the largest proportional increase in female faculty. From Figure 2 one can conclude that much of the increase is the result of hiring female assistant professors. Group Public 136 2nd is the only one of the four groups of rank 136 for which assistant professors outnumber the other ranks. Somewhat surprisingly, for all three other panels representing institutions of rank 136, the average number of female full professors outnumbers the other ranks. In the panels representing institutions of rank 37+, differences in the averages for different rank are small, with assistant professors generally having the largest averages.
Not surprisingly, the majority of the departments have increased the number of female faculty by hiring assistant professors. The question of how the pool of female Ph.D.'s has changed is addressed in the next section.
The number of Ph.D.'s awarded in the U.S. has seen a significant increase since 2002. According to the Taulbee Survey, U.S. Ph.D. production doubled from 2002 to 2008, from about 700 to 1,400. The proportion of females receiving a Ph.D. has increased. For all 163 departments, the increase in proportion went from 15.28% in 1999 to 19.96% in 2009. Due to the increase in overall number and proportion, the number of females receiving a Ph.D. has more than doubled since 2002. Figure 3 shows the proportion and the difference from the baseline average for each of the eight groups.
The proportion of Ph.D.'s awarded to females exhibits some similarities with the trends seen for female faculty. In particular, the figure shows an upward trend for all groups. Although there are differences between the groups, there does not appear to be a group that does significantly better (or worse) than the average. Not surprisingly, the lines representing the proportions are more irregular as the number of Ph.D.'s generated each year generally fluctuates. However, the proportions in the two panel rows in Figure 3 appear to exhibit higher proportions for the groups in 37+. Indeed this is the case. For example, in 2009, the representation of females receiving a Ph.D. is 22% for departments in group 37+ and 18% for departments in group 136.
The proportion of Ph.D.'s awarded to females exhibits some similarities with the trends seen for female faculty.
Figure 4 shows the total number of Ph.D.'s awarded (to male and female students) in departments in group 136 (orange solid) and the total number awarded in departments in group 37+ (orange dashed). The departments in groups 1-36 and 37+ produce about the same number of Ph.D's. In 2009, the 127 departments in group 37+ produced a total of 687 Ph.D.'s and 36 departments in group 136 produced a total of 650 Ph.D.'s. As expected, the average number of Ph.D.'s produced by a department in group 136 is significantly higher; it is 21 Ph.D.'s per year versus eight for departments in group 37+. However, the set of departments in group 37+ awards more Ph.D.'s to females in total numbers as well as proportions.
Figure 4 also shows total Ph.D. production versus faculty hiring. The blue lines show faculty hired by departments in the groups 1-36 (solid) and 37+ (dashed). The data supports what many departments have experienced: since 2002, the number of new Ph.D.'s seeking an academic position has dramatically increased while the number of tenure-track faculty hired has decreased. The peak of the dashed purple line in 2008 represents a hiring spurt in group 37+; the full report shows that it happened in Public 37+ Large.
Ph.D.-granting institutions in the U.S. generate only a fraction (less than 20%) of the undergraduate degrees in computer science; however, trends found in the data for Ph.D.-granting institutions are typically also present in data for all bachelor's degrees. Such data is available from WebCASPAR (https://webcaspar.nsf.gov/). Since 2003, there has been a drop in total bachelor's degree production, both for both male and female students. In addition, there has been a decline in the proportion of females receiving B.S. degrees: from 18% in 2000 to 10% in 2008.
Figure 5 shows the total numbers for undergraduate degree production for males and female students for public versus private institutions. The data suggests a peak in the number of bachelor's degrees awarded between 2003 (for females) and 2004 (for males). The total number of degrees awarded in the public schools is considerably larger than that in the private schools, but the average counts are not so disparate (see the full report for supporting plots).
Figure 6 shows the proportions of female students receiving a bachelor's degree in private and public institutions, with the left panel showing the proportions for all institutions ranked 136 and the right panel for institutions 37+. The declining trend in proportions starts at different times, with public institutions experiencing it earlier than the private ones. Since 2004, the private institutions in group 136 have an increasingly higher percentage of female undergraduates than public ones. The relatively low response rate to the Taulbee Survey from departments in group Private 37+ most likely impacts the irregular shape of the corresponding line. The full report indicates B.S. enrollment in Ph.D.-granting departments appears to have stabilized, but no significant overall increases can be observed.
We have examined the last 10 years of Taulbee Survey data to investigate claims about gender representation in U.S. computer science departments, at all stages of the pipeline. The data shows increases in both female faculty (tenured and tenure-track) and female Ph.D. graduates, but a decline in females receiving B.S. degrees. In aggregate, these findings are consistent with current views of the state of CS. However, surprisingly, our analysis also shows the changes are not uniform with respect to department type. This leads to a number of interesting questions.
First, there are two questions related to the representation of females: Why do many top-ranked departments have so few female assistant professors? Why do lower-ranked departments graduate a higher percentage of female Ph.D.'s? Second, the analysis highlights the groups that have successfully improved gender representation: What strategies did departments in Public 136 2nd have to successfully hire female faculty? Did they attract more female applicants or did they approach the process of hiring differently? What strategies did Private 136 departments use to increase the proportion of female undergraduates? Did they attract new students to CS or did they attract existing CS students to their institution?
Future studies of trends should differentiate Ph.D.-granting departments on more dimensions than just rank in order to gain a better understanding of gender demographics.
Our overall analysis illustrates that future studies of trends should differentiate Ph.D.-granting departments on more dimensions than just rank in order to gain a better understanding of gender demographics.
We thank Betsy Bizot from CRA for generating the underlying data sets. Without her support and guidance in understanding the data, the report on which this Viewpoint is based could not have been developed.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2011 ACM, Inc.