Tracy Camp



The Incredible Shrinking Pipeline

Tracy Camp1
Colorado School of Mines
An edited version of this paper appears in Communications of the ACM, vol. 40, no. 10, pp. 103-110, Oct. 1997.

Many of us in computer science are aware of the pipeline shrinkage problem; the pipeline represents the ratio of women involved in computer science from high school to graduate school. In this article, we discuss the incredible shrinking pipeline. In addition to the pipeline shrinking from high school to graduate school, the pipeline also shrinks at the bachelor's level. Furthermore, while the percentage of bachelor's degrees awarded in CS to women decreased almost every year over the last decade, the corresponding percentages of other science and engineering disciplines increased. Since the number of women at the bachelor's level affects the number of women at levels higher in the pipeline and in the job market, these facts are of great concern. In this article, we look at the harsh facts concerning the percentage of degrees awarded in CS to women since 1980 and we speculate on what the future holds. Lastly, we request the community to respond to the issues presented in this article.

Categories and Subject Descriptors: K.3.2 [Computer and Information Science Education]:
Computer Science Education; K.4 [Computers and Society]: Social Issues

General Terms: women in computing, pipeline shrinkage, computer science education




Introduction

The pipeline shrinkage problem concerning women in computer science is a known phenomenon; we illustrate the phenomenon in Figure 1. Although women make up 50% of high school Computer Science (CS) classes[Walker and Rodger 1996], the percentage of bachelor's degrees in CS awarded to women in the 1993-94 academic year was only 28.4%[NCES 1996].2 At  the graduate level, for the academic year 1993-94, the percentages of degrees in CS awarded to women dropped even further: 25.8% at the M.S. level and 15.4% at the Ph.D. level. In addition, for women who become faculty members, the pipeline shrinks through the academic ranks. According to the CRA Taulbee Survey, only 15.6% assistant professors, 9.4% associate professors, and 5.7% full professors were women in CS Ph.D.-granting departments during the academic year 1993-94 [Andrews 1994-97].

The title of this paper, however, is the incredible shrinking pipeline: the percentage of bachelor's degrees awarded in CS to women decreased almost every year over the last decade. In other words, not only does the pipeline shrink from high school to graduate school, but it also shrinks at the bachelor's level. (See Figure 2.) Furthermore, while the percentage of bachelor's degrees awarded in CS to women decreased, corresponding percentages of other science and engineering disciplines increased. Since the number of women at the bachelor's level affects the number of women at levels higher in the pipeline and in the job market, these facts are of great concern.

There are a number of reasons why we need to improve the percentage of degrees awarded in CS to women. We direct readers interested in these reasons to [Pearl et al. 1990]. In short, there is a critical labor shortage in CS and, although women are more than half the population, they are a significantly underrepresented percentage of the population earning CS degrees.

In this article, we look at the harsh facts concerning the percentage of degrees awarded in CS to women. First, we study the trend of degrees awarded in CS since 1980. We then compare the trend in CS to other science and engineering disciplines. Second, we consider the relationship between the percentage of degrees awarded to women by a CS department and the college the CS department is within. We find that CS departments in engineering colleges graduate, on average,  proportionately fewer women than CS departments in non-engineering colleges. Third, we predict what the future holds; we speculate on future numbers of degrees awarded in CS and we discuss the effect the future numbers have on the female ratio. Lastly, we request that the community respond to the facts and speculations presented in this article. We will compile the responses of the community and present the results in a future CACM issue.





Trends in Degrees Awarded: 1980-81 through 1993-94

We have already established that as women progress from high school to graduate school, they will be part of a smaller and smaller proportion of students. In this section, we consider the total number of degrees awarded in CS at each level and we further examine another kind of shrinking in the pipeline. From 1983-84 to 1992-93, the percentage of B.A./B.S. degrees awarded in CS to women consistently decreased. The percentages of M.S. degrees awarded in CS to women were more stable over this period, and the percentages of Ph.D. degrees awarded in CS to women (though still meager) increased.


The National Center for Education Statistics at the U.S. Department of Education classifies Computer Science departments within the Computer and Information Sciences (CIS) category. Figure 3 illustrates the percentage of degrees awarded within different fields of study in CIS for B.A./B.S., M.S., and Ph.D. levels in 1993-94. As illustrated, the bulk of the degrees awarded are in the general computer science category, with information science and systems a distant second. We, therefore, use the acronym CS is this article to represent all the fields of study in CIS. Table 1 lists the number of B.A/B.S, M.S. and Ph.D. degrees awarded in CS (i.e., CIS) from 1980-81 to 1993-94 and the percentage of recipients that were women.

Although the percentages of Ph.D. degrees awarded in CS to women are very low, the numbers from the three most recent years available suggest the percentages are in an upward trend. However, due to the shrinking of the pipeline at the B.A./B.S. level (discussed below), it is unlikely that this upward trend will continue. From Table 1, we conclude that there is some good news at the Ph.D. level.

Table 1. Degrees awarded in CS: 1980-81 to 1993-94 
Academic Year
B.A./B.S. Degrees 
 % Women
M.S. Degrees
% Women 
Ph.D. Degrees 
% Women
1980-81 
15,121
32.5
4,218
23.0
252
9.9
1981-82 
20,267
34.8
4,935
26.5
251
8.4
1982-83 
24,510
36.3
5,321
28.3
262
13.0
1983-84 
32,172
37.1
6,190
29.3
251
10.4
1984-85 
38,878
36.8
7,101
28.7
248
10.1
1985-86
41,889
35.7
8,070
29.9
344
13.1
1986-87
39,589
34.7
8,481
29.4
374
13.9
1987-88
34,523
32.4
9,197
26.9
428
11.2
1988-89
30,454
30.8
9,414
28.0
551
15.4
1989-90
27,257
29.9
9,677
28.1
627
14.8
1990-91
25,083
29.3
9,324
29.6
676
13.6
1991-92
24,557
28.7
9,530
27.8
772
13.3
1992-93
24,241
28.1
10,163
27.1
805
14.4
1993-94
24,200
28.4
10,416
25.8
810
15.4

Some members of our community are concerned about the first point above. As the number of Ph.D. degrees awarded in CS continues to increase, university positions are more difficult to obtain. Furthermore, many recent Ph.D. graduates do not seem prepared for the realities they face in the job market [Chandra et al. 1994].3 In contrast to this viewpoint, we present the following numbers. First, the unemployment rate for people with a Ph.D. in CS is extremely low (1.1%) [Keaton and Hamilton 1996].4 Second, the underemployment rate for people with a Ph.D. in CS, or the rate of people with a Ph.D. in CS who work outside the CS discipline, is low (3.6%) [Keaton and Hamilton 1996]. Third, there is only a 4% Ph.D. surplus in CS, which is "the lowest surplus of any science and engineering field." [Keaton and Hamilton 1996].

There is mixed news at the M.S. level as well. The number of M.S. degrees awarded in CS continues to increase, but the percentage of M.S. degrees awarded in CS to women has reached its lowest level since 1980-1981. Due to the shrinking of the pipeline at the B.A./B.S. level (discussed below), it is likely that the percentage of M.S. degrees awarded in CS to women over the next few years will continue to decline.

At the B.A./B.S. level, there is only bad news

The fact that the number of B.A./B.S. degrees awarded in CS is decreasing at a faster pace for women than men is especially striking when one considers the total population receiving B.A./B.S. degrees, and the percentage of bachelor's degrees awarded to women by disciplines similar to CS.

Table 2. B.A./B.S. degrees awarded in all disciplines: 1980-81 to 1993-94
Academic Year
B.A./B.S. Degrees
% Women
1980-81
1981-82
1982-83
1983-84
1984-85
1985-86
1986-87
1987-88
1988-89
1989-90
1990-91
1991-92
1992-93
1993-94
935,140
952,998
969,510
974,309
979,477
987,823
991,264
994,829
1,018,755
1,051,344
1,094,538
1,136,553
1,165,178
1,169,275
49.8
50.3
50.6
50.5
50.7
50.8
51.5
52.0
52.5
53.2
53.9
54.2
54.3
54.5

Table 2 lists the total number of B.A./B.S. degrees awarded in all disciplines from 1980-81 through 1993-94, and the percentage of the recipients that were women. The number of B.A./B.S. degrees awarded in all disciplines continues to increase, and the increase is at a faster pace for women than for men.

Figures 4 and 5 compare the trend of degrees awarded in CS with the trends of degrees awarded in disciplines similar to CS from 1980-81 through 1993-94. The majors within a discipline are determined by the National Center for Education Statistics:

Figure 4 illustrates that, while the number of B.A./B.S. degrees awarded in other science and engineering disciplines decreased from 1985-86 to 1993-94, the decrease in CS was the most extreme. From 1985-86 to 1993-94, Eng decreased 18.2% (95,660 to 78,225), Math decreased 16.0% (17,147 to 14,396), and Phy decreased 15.3% (21,717 to 18,400), but CS decreased 42.2% (41,889 to 24,200). Bio/Life increased 33.4% (38,524 to 51,383) during the same period. Figure 5 illustrates that the percentages of B.A./B.S. degrees awarded to women by science and engineering disciplines (except CS) increased almost every of the last 13 years. CS is the only science and engineering discipline where the percentage of bachelor's degrees awarded to women decreased. From 1980-81 to 1993-94, Bio/Life increased 16.3% (44.1% to 51.3%), Eng increased 44.7% (10.3% to 14.9%), Math increased 10.0% (42.1% to 46.3%), and Phy increased 36.6% (24.6% to 33.6%), but CS decreased 12.6% (32.5% to 28.4%). From 1983-84 to 1993-94, CS decreased 23.5%.

In summary, even though more women are awarded B.A./B.S. degrees, and even though the percentages of B.A./B.S. degrees awarded to women in disciplines similar to CS increased, the percentage of B.A./B.S. degrees awarded in CS to women decreased.




The College of Engineering Effect

In this section, we consider the relationship between the percentage of degrees awarded to women by a CS department, and the college the CS department is within. We find that CS departments in engineering colleges graduate, on average, proportionately fewer women than CS departments in non-engineering colleges.

We derived the results given in this section in the following manner. (See [Camp 1997] for more details.) The National Center for Education Statistics provided a list of the number of degrees awarded in CS for two academic years: 1991-92 and 1992- 93. This list was itemized by gender, institution, and degree. A CS department within a university was classified in either an engineering college (CoE) or a nonengineering college (A&S), based on the description of the university within the 1993 Peterson's Guide to Graduate Programs in Engineering and Applied Sciences. (Only universities that offer graduate degrees in CS were considered, since most universities without a graduate program in CS are within A&S.) Although we were not able to classify every department, and although we only considered B.A./B.S. degrees from departments that offer M.S. or Ph.D. degrees, we were able to classify a large number of the CS degrees awarded in 1991-92 and 1992-93 into either CoE or A&S. For example, we were able to classify 47.1% of all B.A./B.S. degrees awarded in CS in 1991-92 (11,561 of 24,557). These 11,561 degrees were awarded from 263 CS departments. (See Table 3 for further details on the data classified in this study.) Note that we only considered degrees awarded in CS; we did not include degrees awarded in  Computer Engineering (CE). (The percentage of degrees awarded in CE to women is dramatically lower than the percentage of degrees awarded in CS to women. For example, only 11.7% of the B.A./B.S. degrees awarded in CE went to women in 1992-93.)

Table 3. Data classified in this study
Academic Year B.A./B.S. Degrees Classified M.S. Degrees Classified Ph.D. Degrees Classified
1992-93 11,561 (47.1%) 
from 263 depts.
7,139 (74.9%) 
from 243 depts.
671 (86.9%) 
from 86 depts.
1992-93 11,591 (47.8%) 
from 260 depts.
7,618 (75.0%) 
from 241 depts.
691 (85.8%) 
from 94 depts.

The results are displayed in Table 4. The table contains the number of degrees classified into either CoE or A&S at the B.A./B.S., M.S., and Ph.D. levels, and the percentage of the recipients that were women, for the two academic years analyzed. At both the B.A./B.S. and M.S. levels, if a CS department moves from A&S to CoE, the percentage of degrees awarded to women decreases by approximately 18~26%. At the Ph.D. level, the decrease is less pronounced; if a CS department moves from A&S to CoE, the percentage of Ph.D. degrees awarded to women decreases by approximately 7-8%.

Table 4. The College of Engineering Effect
Academic Year B.A./B.S. Degrees  % Women M.S. Degrees % Women Ph.D. Degrees  % Women
1991-92/CoE 
1991-92/A&S
3,633 
7,928
22.5 
28.3
2,613 
4,526
23.6 
29.2
360 
311
11.7 
12.5
1992-93/CoE 
1992-93/A&S
3,710 
7,881
22.8 
26.8
3,082 
4,536
23.4 
28.2
351 
340
12.2 
13.2

The results shown in Table 4 suggest that CS departments in engineering colleges, on average, graduate proportionally fewer women than CS departments in nonengineering colleges. We performed statistical analysis on the percentage of women graduates for the 1991-1992 academic year. We found that the difference between the percentage of B.A./B.S. degrees awarded to women by CS departments in engineering colleges and the percentage of B.A./B.S. degrees awarded to women by CS departments in non-engineering colleges was statistically significant with the traditional cutoff of a = 0.01. The difference for bachelor's degrees was significant both on the raw percentage data and when the percentages were transformed using the arcsin transformation [Steel and Torrie 1980]. The difference in the M.S. degrees awarded was significant at a = 0.2. The statistical analysis strongly suggests that the proportion of females graduating with B.A./B.S. degrees is significantly lower when a CS department is within an engineering college. (In other words, it is highly unlikely that the observed difference is a random event.) The statistics do not allow a similar confidence in the significance of the M.S. data.

Over the last decade, we are aware of a number of universities that have moved their CS department from a non-engineering college to an engineering college (e.g., Pennsylvania State University and University of Washington). If CS departments continue to move to engineering colleges, we may continue to see a decrease in the percentage of B.A./B.S. degrees awarded in CS to women. There are, however, other factors that may affect the percentage of B.A./B.S. degrees awarded in CS to women as well.




Future Predictions: 1994-95 through 2006-07

The computer industry is growing rapidly. There is a critical shortage of computer scientists in today's job market—nearly 190,000 unfilled information technology positions, not counting small business, government, and nonprofit employers, exist in the United States alone [Arnheim 1997]. Furthermore, in the year 2005, it is predicted that the number of computer professionals employed as computer scientists and system analysts will be almost double what the numbers were in 1994[Keaton and Hamilton 1996]. In 1994, there were more than 149,000 and 482,000 computer scientists and system analysts, respectively, in the work force. In 2005, it is predicted that there will be more than 282,000 and 927,000 computer scientists and system analysts, respectively, in the work force. As Table I shows, we have only been graduating approximately 35,000 students with degrees in CS each year. Thus, it is critical that the number of B.A./B.S. degrees awarded in CS over the next decade increases, instead of continuing on its current decreasing trend.

In this section, we attempt to predict the future, and we have some good news. There is positive evidence that the number of B.A./B.S. degrees awarded in CS will increase in the near future. We speculate on how this increase will affect the percentage of women recipients.

Table 5 lists the projected total number of B.A./B.S. degrees awarded in all disciplines over the next decade, and the projected percentage of the recipients that will be women [NCES 1997]. The National Center for Education Statistics predict that the number of B.A./B.S. degrees awarded at the turn of the century will be slightly fewer than the number that is currently being awarded. However, the number of B.A./B.S. degrees awarded in a decade from now will overcome this turn of the century decrease, as well as significantly surpass the number that is currently being awarded. Furthermore, while the number of B.A./B.S. degrees awarded will fluctuate over the next decade, the percentage of B.A./B.S. degrees awarded to women will continue on its upward trend.

Table 5. Future predictions of B.A./B.S. degrees awarded in all disciplines: 1994-95 to 2006-07
Academic Year
B.A./B.S. Degrees
% Women
1994-95
1995-96
1996-97
1997-98
1998-99
1999-00
2000-01
2001-02
2002-03
2003-04
2004-05
2005-06
2006-07
1,181,000
1,186,000
1,183,000
1,169,000
1,140,000
1,138,000
1,151,000
1,169,000
1,191,000
1,216,000
1,237,000
1,253,000
1,268,000
54.9
55.2
55.4
56.5
56.1
56.0
56.2
56.5
56.7
57.0
57.6
57.8
58.0

Although the total number of degrees awarded at the B.A./B.S. level is expected to decrease over the next five years, we should see a dramatic increase, not a further decrease, in the number of B.A./B.S. degrees awarded in CS. According to the 1996 CRA Taulbee Survey, the number of B.A./B.S. degrees awarded in CS should increase in the near future, as the number of new bachelor students enrolled in computer science Ph.D.-granting departments increased 40% in fall 1996 [Andrews 1994-97). Since the number of B.A./B.S. degrees awarded in CS by Ph.D.-granting departments should dramatically increase in the near future, we expect to see a corresponding increase in the number of B.A./B.S. degrees awarded in CS by non Ph.D.-granting departments.

One question, however, remains: how will the increase in the number of B.A./B.S. degrees awarded in CS affect the percentage of women recipients? As listed in Table I, the decrease in the number of B.A./B.S. degrees awarded in CS over the last decade was at a faster pace for women than for men. Will the increase in the number of B.A./B.S. degrees awarded in CS occur at a faster pace for women? Or will the proportion of women continue to lag behind their male colleagues? The 1996 CRA Taulbee Survey does not classify the 40% increase of bachelor students in CS by gender [Andrews 1994-97], so we can only speculate on what the outcome will be.

First, we consider why there is such an increase in the number of new bachelor students in CS. The 1996 CRA Taulbee Survey suggests that the marketplace and the Web contributed to this increase [Andrews 1994-97]; the job opportunities in CS are becoming well-known, and the Web has brought more attention to our discipline than anything in the past.

Second, let us speculate on how the Web will affect the percentage of B.A./B.S. degrees awarded in CS to women. The seventh study of Web users at the Graphic, Visualization, fe Usability Center (GVU) at Georgia Institute of Technology states that the core demographics of Web users are stabilizing[GVU 1997]. In other words, the rate of change in the demographics is much slower than in previous years, e.g., the percentage of female respondents to the seventh survey is almost exactly the same as the percentage of female respondents to the survey the previous year. In the seventh survey, with almost 20,000 unique respondents, 31.3% of the responses were from females. While other Web surveys are available, which claim a higher proportion of female respondents, we find the GVU survey is appropriate in our speculation. The GVU survey claims that, by the nature of their sampling method, the respondents represent active Web users. If the Web has affected the increase in new students enrolled at the bachelor's level, we believe it is the active users that are contributing to this effect.

The percentage of B.A./B.S. degrees awarded in CS to women in 1993-94 was 28.4%. Since the female ratio of active Web users is slightly higher (31.3% versus 28.4%), we may see a slight increase in the percentage of degrees awarded in CS to women in the near future. This slight increase, however, is much lower than the 37.1% we enjoyed in 1983-84.




Community Response

Why has the number of B.A./B.S. degrees awarded in CS decreased since 1985? Why is the decrease occurring at a faster pace for women than men? While the number of B.A./B.S. degrees awarded in CS in 1993-94 is almost equivalent to the number of B.A./B.S. degrees awarded in CS in 1982-83 (24,200 versus 24,510), why is the percentage of degrees awarded to women dramatically smaller today (28.4% versus 36.3%)? The percentage of B.A./B.S. degrees awarded to women in all disciplines increased from 1980-81 to 1993-94; the percentages of B.A./B.S. degrees awarded to women in disciplines similar to CS increased over the same time period. Why has the percentage of B.A./B.S. degrees awarded to women in CS decreased? Why is the percentage of B.A./B.S., M.S., and Ph.D. degrees awarded to women by CS departments in engineering colleges smaller than the corresponding percentages in non-engineering colleges? Will the percentage of degrees awarded in CS to women continue to lag behind in the next decade? And, most importantly, what can we, as a community, do to improve the situation?

In a paper on women in science and engineering, John White, Dean of Engineering at Georgia Institute of Technology, said, "If we want a different outcome, we're going to have to do things differently. We're making too little progress doing more of the same thing. The time for evolution is passed; it's time for revolution." [White 1995] White requests a revolution to improve on the small amount of progress in attracting and retaining more women in science and engineering over the last decade. Instead of progress, however, we have seen a deterioration in CS. In order to eventually make progress, computer scientists and educators seem to need an especially dramatic change in direction. We hope the CS community will become involved in  exploring the options and planning changes. To that end, ACM-W (the ACM Committee on Women in Computing) has established a web site, ACM-W , which includes data, references, and some of the ideas suggested for improving the current situation. There is also an interactive survey available at the web site; we request that the CS community considers the above questions and responds to the survey. We will compile the responses of the community and present the results,  including strategies proposed for attracting and retaining women in CS, in a future CACM issue. If we work together, perhaps we can identify and implement the changes that are necessary to reverse the alarming decline of women's participation in CS.




Acknowledgments

We gratefully acknowledge Keith Miller for his contributions on the statistical analysis presented in this paper. We also thank Anita Borg, Denise Gurer, Janie lrwin, Keith Miller, and Glen Oberhauser for providing helpful suggestions that improved the quality of this paper. Lastly, we thank Nancy Deanne Moore and Keely Wilson for their help in classifying the data presented. This work was supported in part by NSF Grant NCR-9702449.




References

ANDREWS, G. 1994-97.  CRA Taulbee surveys. Computing Research News 6-9, 2 (March).
ARNHEIM, L. 1997.  IT companies heavily courting CS grads. Computing Research News 9, 3 (May), 1-7.
CAMP, T. 1997.   CS programs in engineering colleges = fewer females. Journal of Women and Minorities in Science and Engineering. Volume 4, No. 1, pp. 15-25, January 1998.
CHANDRA, A., PATTERSON, D., TRAUB, J., AND YOUNG, P. 1994.  Do we produce too many Ph.D.s? Computing Research News 6, 3, 1-2.
GVU. 1997.   GVU's 7th WWW user survey. Graphic, Visualization, fe Usability Center, Georgia Institute of Technology, Survey .
KEATON, J. AND HAMILTON, S. 1996.   Employment 2005: Boom or bust for computer professionals? IEEE Computer 29, 5, 87-98.
NCES. 1996.   Digest of Education Statistics. National Center for Education Statistics, U.S. Department of Education, Washington, D.C.
NCES. 1997.   Projections of Education Statistics to 2007. National Center for Education Statistics, U.S. Department of Education, Projections .
PEARL, A., POLLACK, M., RISKIN, E., THOMAS, B., WOLF, E., AND Wu, A. 1990.  Becoming a computer scientist. Communications of the ACM 33, II (November), 47-57.
STEEL, R. AND TORRIE, J. 1980.   Principles and Procedures of Statistics, a Biometrical Approach: 2nd Edition. McGraw-Hill.
WALKER, E. AND RODGER, S. 1996.  PipeLINK: Connecting women and girls in the computer science pipeline. In Proceedings of the National Education Computing Conference (1996), pp. 378-384.
WHITE, J. 1995.   Women in engineering and science: Does anyone care? In Proceedings of Bridging the Gender Gap in Engineering and Science.: The Challenge of Institutional Transformation (1995), pp. 41-48.



1 Tracy Camp is a member of the ACM Committee on Women in Computing (ACM-W). The mission of ACM-W is to engage in activities that aim to improve computing environments in order to gain equity for women in computing.

2 The number of degrees awarded for the academic years 1994-95 and beyond are not yet available from the National Center for Education Statistics at the U.S. Department of Education.

3Faculty members have a responsibility to disseminate information about the job market to their students, thus allowing students to make informed decisions.

4The two most recent CRA Taulbee surveys cite the unemployment rate of new Ph.D. graduates at 1.7% in 1995 and 0.8% in 1996 [Andrews 1994-97]. Some claim the unemployment rate of the Taulbee surveys may be misleading [Chandra et al. 1994]; for example, the 1995 and 1996 surveys do not know what 18.5% and 15.2%, respectively, of the Ph.D. graduates are doing. The unemployment rate from [Keaton and Hamilton 1996], however, is based on a report from the Survey of Doctorate Recipients conducted by the National Science Foundation. Since the two unemployment rates are comparable, we believe the rate is sound.



Tracy Camp received her Ph.D. from the College of William and Mary and is currently an associate professor at The Colorado School of Mines. Her principal research interests are mobile computing and  networking.
Address: Department of Mathematical and Computer Sciences, The Colorado School of Mines, Golden, CO 80401,  email: tcamp AT mines DOT edu