[Published originally in the March 2003 edition of Computing Research News, Vol. 15/No. 2, pp. 2-3.]
Expanding the Pipeline
Migrating Out of Computer Science
By J McGrath Cohoon and Lih-Yuan Chen
Undergraduate enrollment in the computer science major skyrocketed through the mid- to late 1990s. Resources were strained as the number of students declaring a CS major rose year after year at rates that made it difficult for departments to find the necessary space, equipment, and faculty. Since 2000, this demand for a CS major appears to have leveled off and may even be declining. (For more information on the trends in newly declared majors, see "Survey Results Show Better Balance in Supply and Demand" by Vardi, Finin, and Henderson, elsewhere in this issue of CRN.)
However, the halting growth in numbers of new students is only half the story. There is now evidence that even before the number of students choosing a computing major started to fall, undergraduates who were already in computing majors began migrating out at higher rates than in the past. In other words, not only are fewer students going into a computing major, but more are switching out before earning their degree. This phenomenon occurs at varying rates in different types of programs.
Data from a new nationwide study of computer science programs reveal recent trends in both enrollment and migration out of the major. This study, "Departmental Factors in Gendered Attrition from Undergraduate IT Majors," focused on mid- to large-sized undergraduate programs in the contiguous United States. For this portion of the study, data came from 71 institutions. Ninety percent of these institutions were public, and 60 percent of the study institutions offered doctoral programs. We report here on the migration of all CS students. Analyses of gender differences will be discussed in other reports.
The Departmental Factors data in Table 1 show the annual rise in the study departments' total undergraduate enrollment. From 1994 to 1999, CS undergraduate enrollment at the average institution in this study grew by approximately 100 percent. These data show the recent changes in total student enrollment numbers, not new student numbers. The latter phenomenon is apparent in data from the 2000-01 and 2001-02 Taulbee Surveys of Ph.D-granting institutions, which produced similar, but not identical, data to that produced by the Departmental Factors study. The Taulbee Survey measured newly declared majors and degrees awarded, not total enrollment.
More importantly, Figure 1 shows that while the average undergraduate enrollment was rising dramatically, students who remained at their institutions migrated out of the CS major at an average rate of 16 percent per year. (Note that the figure shows this number split into Ph.D. and non-Ph.D.-granting institutions and by upper and lower level.) During the years when enrollments increased, the trend in overall attrition rates declined from 16 percent in 1994 to a low of 14 percent in 1996. But attrition began to grow in 1998 and reached a high of 19 percent in 1999, even before new enrollment stopped rising. This increase in attrition rates at the end of the 1990s shows that students began migrating out of the CS major even before fewer came into the major.
The migration out was not an effect of students choosing to enter the job market early. We know this because our migration rates only consider students who did not leave their institution. They continued as undergraduates, but in a different major. More likely causes are changes in the job market that led students motivated by career options to choose a different major, or growing student disaffection with conditions in departments that lacked adequate resources. Migration out is also not likely to be extreme relative to other science, mathematics, and engineering disciplines. Although no current data are available for comparison, Seymour and Hewitt (1997) showed that students leave the CS major at relatively low rates compared with these other disciplines.
The trends shown by the Departmental Factors data were similar for both Ph.D.-granting and non-Ph.D.-granting institutions. They differed only in that the average Ph.D. institution saw more students migrate out of computer science than did the average non-Ph.D.-granting institution. This difference in CS attrition rates for the different types of institutions continued until the 1999-2000 academic year when there was a large increase in attrition, particularly at non-Ph.D.-granting institutions (see Figure 1).
Not surprisingly, attrition was highest among first- and second-year students. Freshmen and sophomores migrated out of the computer science major each year at an average rate of 19 percent. Juniors and seniors were much less likely to leave for a different major. Their average attrition rate was only 12 percent. The rate for upper-level students was less in non-Ph.D.-granting institutions than in Ph.D.-granting institutions, and it demonstrated less of an increase in 1999 than did lower-level student attrition (see Figure 1).
Computer science departments varied considerably in their attrition rates (Figure 2), ranging from 1 to 66 percent. Departments in Ph.D.-granting institutions had rates that ranged from 1 percent to 53 percent. Departments in non-Ph.D.-granting institutions had rates that ranged from 5 to 66 percent. This variation in attrition rates suggests that some departmental characteristics might influence the rates at which students leave their programs.
Several of the characteristics that one might expect to influence departmental attrition rates were investigated for their relationship with attrition. Thus far, we have no evidence that students were any more or less likely to leave the major in public institutions than private institutions. Likewise, we found no relationship between attrition rates and student/faculty ratio. Looking at freshmen, sophomores, juniors, and seniors together, none of the measures of academic quality we considered was significantly related to overall attrition rates.
However, it appears that upper- and lower-level students were motivated to leave by different factors, although there was a tendency for departments with high attrition among one group to also have high attrition among the other group (r= .43, significant at the .01 level). For juniors and seniors, we have only conducted preliminary analyses and have no results to report yet. For lower-level students, we found two factors that significantly correlated with attrition when we considered upper- and lower-level students separately.
The two significant correlations with attrition rates for lower-level students were overall grade point average (GPA) for a department's lower-level students and faculty scholarly quality at Ph.D.-granting institutions. Freshmen and sophomores were more likely to leave departments where the average overall grade point average was low (r=-.47, significant at the .01 level), and at Ph.D.-granting institutions, lower-level students were also more likely to leave departments where the National Research Council rating of faculty scholarly quality was low (r=-.62, significant at the .01 level).
The measure of departments' faculty scholarly quality was produced by the National Research Council in 1993 by averaging the ratings of external experts. Ratings are strongly correlated with citations per faculty publication, research productivity, size of program, and production of new Ph.D.s. (Ehrenberg and Hurst, 1996). The rating scale is from zero to five with zero representing "not sufficient for doctoral education" and five representing "distinguished." In our study, the mean for rated departments was 2.44 with a standard deviation of 0.90. The minimum rating received was 1.35, and the maximum was 4.18.
The relationship between faculty scholarly quality and attrition calls to mind findings from research on attrition from science, mathematics, and engineering disciplines in general. Seymour and Hewitt (1997) found that the high rate at which students migrate out of these disciplines was due to factors related to students' collegiate academic experiences. In particular, when they encountered the poor teaching and advising, harsh grading, and heavy demands that are not uncommon in these disciplines, many students switched to a different major.
If faculty scholarly quality and these pedagogical practices are related, it would explain why students are more likely to leave higher-rated departments. Our survey of CS faculty showed that faculty in departments rated high in scholarly quality were less likely to consider the lecture the most important element of their instruction (r=-.58, sig. at .01 level), and less likely to consider their responsibility as instructors satisfied by presenting information to the students (r=-.47, sig. at .05 level). To the extent that these opinions represent good pedagogical practices, they hint that in Ph.D.-granting institutions, departments with faculty of the highest scholarly quality may also be those with the best pedagogy. This hypothesis receives some support from our focus group data where students reported that knowledge and the ability to communicate it were very important faculty qualifications. When these qualities were lacking, undergraduates were discouraged in their CS studies.
Grading was the other factor that contributed to lower-level students leaving. Departments where the average freshman and sophomore had a low GPA were more likely to lose students than were departments where grades were higher. Partial correlations showed that the effect of grades persisted even when controlling for the median SAT score of an institution's incoming freshmen, and when controlling for the average math SAT of lower-level students in the department. Furthermore, there were no significant relationships between attrition rates and either the median SAT or the math SAT. Thus, the effect of grades on attrition does not appear to be a simple matter of students' academic quality (see Figure 3).
Focus group data suggest that low grades may lead some students to switch majors because students feel they get too little reward for a great deal of effort. For example, numerous students reported that faculty could be overly tough in their evaluation of very difficult work. At least one student argued that receiving low grades in return for extensive, good-quality work was the most obvious reason people left the department.
Computer science departments have been struggling for years to keep up with demand for their programs. However, the data presented here indicate that the steep upward trend in demand has changed direction. New student numbers are falling, and students increasingly are migrating out of undergraduate CS programs.
If interest in the undergraduate CS major continues to decline, departments may have to make recruiting and retaining students more of a priority. Hints at effective intervention points are provided by the departmental and institutional characteristics that vary with student attrition rates. This first look at those characteristics suggests that freshmen and sophomores are more likely to leave the CS major when grade point averages for students in their department are low, and in Ph.D.-granting institutions, when the scholarly quality of their CS faculty is low.
The results discussed here are preliminary; much further investigation is needed before we can confidently assert which factors affect departments' attrition rates. Yet despite the need for more research on predictive factors, the trends demonstrating a decline in undergraduate interest in CS now have more than one data set supporting them. As the discipline catches its collective breath from the rush to keep up with demand, it might do well for some departments to consider how to attract and retain students.
More information about this project and its results is available online at http://curry.edschool.virginia.edu/ITattrit
Joanne McGrath Cohoon is a Research Assistant Professor in the Curry School of Education at the University of Virginia. She is P.I. of the NSF-funded study, "Departmental Factors in Gendered Attrition from Undergraduate IT Majors," which collected data used for this article. Lih-Yuan Chen is a Graduate Research Assistant in the Curry School of Education at the University of Virginia where she expects to earn her Ph.D. in Special Education.
This material is based on work supported by the National Science Foundation under grant number EIA0089959. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Copyright © 2007 Computing Research Association. All Rights Reserved. Questions? E-mail: email@example.com.