Communications of the ACM

Communications of the ACM

Toward Improving Female Retention in the Computer Science Major

By J. McGrath Cohoon
Communications of the ACM, May 2001, Vol. 44 No. 5, Pages 108-114
10.1145/374308.374367
Comments

Although many computer professionals believe that inherent or deeply ingrained gender differences make women less suited to the study and practice of computer science [5, 9], the results reported here demonstrate that female underrepresentation in computer science could be avoided. Women can and do succeed in computer science (CS) when conditions do not deter them. The variation that occurs in women's participation rates demonstrates that many women succeed as computer scientists in certain times and settings. Conditions affecting female retention in undergraduate computer science are identified in this article.¹

Evidence that women's success in computer science varies over time was provided in an article by Camp that appeared in Communications in 1997 [2]. In this article, Camp documented the rise and fall in the female proportion of computer science Bachelor's degrees between 1981 and 1994. Camp also noted that this variation was affected by the type of college (engineering/nonengineering) in which a CS department was located. Figure 1 expands Camp's timeframe to the most recent available data and reconfirms that women's proportion of CS Bachelor's degrees waxes and wanes. As Figure 1 shows, women comprised 14% of CS Bachelor's degrees in the U.S. in 1971; this percentage rose to 37% by 1984, and then dropped 10 percentage points over the subsequent 13 years.

These temporal changes in female representation are not statistical phantoms that can be easily explained away. In particular, they are not attributable to general trends in female educational attainment—women's proportion of all Bachelor's degrees rose steadily from 46% to 56% during this period. Furthermore, women's proportion of non-CS scientific and technical disciplines also rose during this period [7]. The temporal changes in female representation were also not attributable to the effects of newly formed CS departments—a similar rising and falling trend in female representation was evident when a fixed set of CS departments was examined. Nor are the trends in CS due to fluctuations in male computer science enrollment only—both male and female enrollment numbers rose and fell during these years. Unfortunately, women's numbers increased more slowly than men's numbers and decreased more rapidly than men's numbers.

The variation over time strongly suggests that factors other than inherent gender characteristics play an important role in the underrepresentation of women in computer science. However, the question remains, if not female characteristics, then what? One part of the answer may lie in discipline characteristics [8]. Another part of the answer, and the focus of this article, lies in the departmental characteristics that affect whether women are retained in the CS major at rates comparable to the retention rates of men.

This assertion that departmental characteristics affect female retention is supported in part by observing that the disproportionate loss of women from undergraduate computer science varies widely across academic departments. Within the same time period, and regardless of the type of school (engineering/non-engineering) or the program's overall attrition rate, some CS departments successfully retained female undergraduates at similar rates to male undergraduates [3]. Other departments lost women at much higher rates than men were lost.

How is it that some CS departments retained female students at comparable rates to men? The analysis reported here identifies departmental factors associated with equal retention of male and female students, showing that CS departments generally retained women at comparable rates to men when:

• The faculty included at least one woman; was stable; valued, mentored, and supervised female students; enjoyed teaching; and shared responsibility for success with their students;
• The department had above average support from its institution;
• Graduating seniors had access to a strong local job market; and perhaps, high starting salaries; and
• There were sufficient numbers of female students in each class for them to support each other.

Thus, in answer to this article's opening question about the causes of disproportionate female attrition, the research described here found that gender composition, faculty attitudes and behaviors, and institutional context can have a substantial effect on whether women leave the undergraduate computer science major at higher rates than men leave.

Back to Top

A Study of Departmental Characteristics and Practices

The research reported here differs from most previous research on this or the related topic of women in science, mathematics, and engineering [4, 6, 12], in that it included many institutions and it analyzed outcomes for departments rather than individual students. Inclusion of multiple departments means that the findings of this research are more likely to apply to CS departments in general. Taking the department as the unit of study made it possible to identify departmental actions and characteristics that are linked with loss of women. These factors are within the control of higher education institutions.

The dependent variable for this research—average difference between male and female attrition rates, or gendered attrition rate—was calculated with official data from the State Council of Higher Education for Virginia. For each study department, the annual number of declared majors who switched to another major relative to the total number of declared majors was calculated for both sexes. Next, these annual male and female attrition rates were averaged across the academic years 1992–1997, and the average female rate was subtracted from the average male rate. This average difference was a department's gendered attrition rate.

Table 1 lists the male and female and gendered attrition rates for each of Virginia's 23 coeducational CS departments between 1992 and 1997. This data indicates that the difference between male and female rates varied from one department to another —gendered attrition did not occur to the same degree in all CS departments.

In addition, the mean gendered attrition rate in Table 1 shows that in Virginia, attrition among females with declared CS majors was twice as high as the male rate for this time period. As is the case in the rest of the U.S., Virginia institutions lost women from the computer science major at a higher rate than men. (Rounding error accounts for cases where the gendered attrition rate differs slightly from the male attrition rate minus the female attrition rate.)

The 23 departments studied represented the entire population of Bachelor's degree-granting coeducational computer science departments in Virginia between 1992 and 1997. (A department is defined as coeducational if it enrolled both male and female students.) Probability theory does not support generalizing data from a single state to the entire U.S., therefore further study is needed with a national sample. However this data documents what occurred at all the institutions in one state, so it may offer insight into the general process of gendered attrition from computer science education. Figure 2 shows that degree trends in Virginia followed those for the U.S. from 1985 through 1995 with only slightly more volatility. Thus, it is reasonable to assume that conditions in Virginia were similar to those in the nation as a whole and that findings from this study might extend beyond the one state in which it was conducted.

To strengthen confidence in the results of this study, it employed three research methods to form a consensus of findings. The methods employed were: qualitative analysis of interview data; discipline comparisons of quantitative survey data; and statistical analyses of CS survey data. As explained here, conclusions were based on results produced with at least two types of analyses.

The first research method involved 34 face-to-face interviews at five of the CS study departments. The five departments interviewed represented CS departments of varying size, geographic location, institutional selectivity, and gendered attrition rates. Interviews were conducted with the chairperson, all of the faculty in small departments and up to five faculty in larger departments, and an average of 11 students per department. For the most part, students were interviewed in gender-segregated groups. The interviews lasted between 45 minutes and two hours during which the realities of undergraduate computer science education and potential explanations for disproportionate female attrition were explored.

The interview data portrayed computer science as a difficult, time-consuming major in which students persist because they enjoy the thrill of creating a working program and because they anticipate a highly rewarded career. The interview data also indicated several factors that could contribute to gendered attrition from CS departments. These factors are identified in Table 2 and are discussed here with the results of the other analyses.

Based on the interview results, a written survey was created and sent to all 23 of the qualifying CS departments. This survey targeted faculty and chairpersons at all of the Virginia CS departments and at the 23 biology/life sciences departments at the same institutions. Up to 10 faculty members plus the chairperson at each department were sent this written survey. A total of 255 completed questionnaires were returned. The overall response rate for all questionnaires was 68%, with a breakdown of 90% of chairpersons responding and 65% of faculty responding.

The data from this written survey was statistically analyzed to determine which factors had the strongest, most consistent effects on gendered attrition. First, descriptive data from computer science and biology/life sciences was compared. Second, quantitative measures of CS department characteristics and practices were correlated with the departments' gendered attrition rates. Third, the characteristics and practices that were significantly correlated with gendered attrition were entered into multiple regressions that compensated for differences in the size of study departments.

The purpose of the discipline-level comparison between CS and biology/life sciences was to further identify factors related to gendered attrition rates. Biology and CS are two scientific/technical disciplines sharing many characteristics and requirements. Yet, biology now enrolls and retains much higher proportions of women than does CS. The data in this study shows that Virginia's average gendered attrition rate for Biology/Life Science was -1% during the study period, meaning that the female attrition rate in Biology/Life Science was only 0.01 greater than the male attrition rate. This rate was much lower than the average gendered rate in Virginia's CS departments, which was -9%. In addition, there was less variation in gendered attrition rates across Virginia's Biology/Life Science departments (Standard Deviation = 0.035) than in computer science (Standard Deviation = 0.148).

The results of both the discipline comparisons and the statistical analyses are presented in Table 2. Together with the interview results, the consensus from these three different analyses shows that faculty characteristics and practices, institutional and community environment, and the availability of same-sex peer support all influenced gendered attrition from the CS major.

Back to Top

Departmental Factors Related to Gendered Attrition

The evidence in support of each finding is discussed in more detail in the following paragraphs.

Gender composition. Detailed qualitative and quantitative analyses of the interview and survey data determined the single strongest factor and most consistent finding of this gendered attrition study was the departments with higher female proportions of enrollment were more likely to retain women at comparable rates to men.

This finding echoed a result from an empirical study of enrollment and retention in the sciences. Astin and Astin [1] found the more socially similar peers students had in their major, the more likely students were to persist in a science, mathematics, or engineering major. The research reported here shows this effect holds in computer science—the presence of female peers helped CS departments retain women at equivalent rates to men.

All three methods of analysis confirmed the effect of female proportion of enrollment on gendered attrition. The discipline comparison showed that in biology, which retained women at comparable rates to men, 58% of the majors were female. In comparison, in CS only 28% of the majors were female. Furthermore, correlations and regressions showed a large portion of the variance in the gendered attrition rates of CS departments was explained by the gender composition of enrolled students at the beginning of the study period.

The interview data offered an explanation for this effect by documenting the crucial role that peer support plays in the success of computer science students. Both male and female students in every interviewed department explained that their classmates were a necessary source of help. As one male senior put it, "To make the classes easier, it's very important to know other people in the classes... . It would be really hard if I didn't know anyone in the class and I couldn't ask anyone a question." Another male student offered evidence that peer support was essential by declaring that he earned his lowest grades in classes where he knew no one.

In departments where there were too few females for same-sex support in CS classes, women had to rely on their male classmates. One female senior who persisted in the CS major despite being the only female in many of her classes spoke of the teasing she encountered when getting help from male classmates. "The guys that hang out with me, they're like, 'You're not very good because you're a girl.'" She interpreted this male behavior as "just picking with me. I don't think they think that." However, if other, more sensitive female students encountered similar teasing from male classmates, it could explain why this department lost female CS majors at a disproportionately high rate. (The gendered attrition rate in this department was -19%.)

Faculty characteristics and practices. In addition to same-sex peer support, several faculty characteristics and practices affected gendered attrition rates. These factors included: faculty turnover or stability; faculty attitudes; mentoring of students; teaching; and the presence of female faculty.

Faculty turnover. The interviews did not address the issue of faculty turnover, but the survey data did provide evidence of a relationship between faculty stability and gendered attrition. The discipline comparison showed that computer science lost faculty at a higher rate than biology lost faculty during the five-year study period. The average CS department's faculty turnover rate was five percentage points higher than the average biology rate. Correlations and regressions also confirmed that CS departments with high faculty turnover were likely to lose women at a disproportionately high rate.

Faculty attitudes. Faculty attitudes toward female students could also influence gendered attrition. The interview departments where most of the faculty expressed strong appreciation for their female students' abilities and work styles were also the only interview departments that retained their women at comparable rates to men. The departments where most faculty reported no difference between their male and female students' abilities, or some female disadvantage, were the departments that lost women at disproportionately high rates. Additional supporting evidence for this finding came from the discipline comparisons. Comparing biology and CS faculty's evaluations of their male and female students confirmed that very positive evaluations of female students' abilities were related to equivalent retention rates for male and female students. However, neither correlations nor regressions of CS data showed significant support for this finding. This lack of statistical support could have been due to the difficulty of assessing such a sensitive issue with a written survey.

Mentoring. Mentoring was defined as "activities such as: recruiting individual students into professional activities; offering personalized advice to individual students; encouraging individual students; and helping individual students establish careers." The time that faculty devoted to mentoring students and the number of students mentored and supervised may also have influenced gendered attrition. Although the interviews produced no conclusive evidence related to mentoring, the quantitative data showed that the more time faculty spent mentoring and the more female students mentored or supervised, the less difference there was likely to be between a department's male and female attrition rates. Regression results showed that regardless of female enrollment, gendered attrition rates improved with both the average number of hours per week faculty mentored students and the number of female students mentored.

Teaching. Faculty approach to teaching was another significant factor in the retention of female students. The importance of teaching to student retention in science, mathematics, and engineering was previously documented by Seymour and Hewitt [10]. In keeping with their results, the current study found that gendered attrition rates were likely to be low in CS departments where faculty enjoyed teaching undergraduates, and where faculty considered student success to be something for which they shared responsibility. The belief in shared responsibility was in contrast to the belief that student success resulted entirely from students' innate abilities.

Every interview department emphasized good teaching, but chairpersons went about enforcing this norm in different ways. The chairperson in a department with low gendered attrition explained his strategy in this way: "You try to figure out what every person does best and try to give them the opportunity to do as much of that as possible." He went on to explain that faculty members who love teaching should teach and those who are "good at research, in a quantifiable way, they ought to be given every opportunity to do that, even if it means reducing their teaching load." This strategy appeared to be successful—students had few complaints about teaching quality in this department and there was little difference between male and female attrition rates.

During the interviews, student complaints about teaching quality varied with department's gendered attrition rates. Students in departments with low gendered attrition rated the teaching they experienced as very good to excellent. In contrast, students in departments with high gendered attrition rates considered their faculty's teaching to be of questionable quality. Students' most frequent specific complaint was about faculty unwillingness to help them master difficult concepts.

The quantitative data confirmed interview findings about teaching. Correlations and regressions showed that Virginia's CS departments were likely to lose women at disproportionately high rates when faculty did not get much personal satisfaction from teaching undergraduates, or when faculty strongly believed that innate ability was responsible for student success. The discipline comparison showed that latter belief was slightly more prevalent in CS than in biology; and its counterpart belief—that faculty shared responsibility for student success—was slightly more prevalent in biology than in CS.

Presence of female faculty. Another faculty characteristic influencing whether female students left the CS major at disproportionately high rates was the presence of female faculty. Departments with no female faculty lost female students at high rates relative to men. The discipline comparison offered support for this finding by demonstrating that biology had fewer departments with no women on the faculty than did CS. However, the average female proportions of faculty were remarkably similar in the two disciplines (27% female in CS, 28% female in biology). Correlations also supported the observation that departments with female faculty retained female students at equivalent rates to male students. Regression results showed a similar relationship between these variables. However, this result was not statistically significant when the gender composition of enrolled students was taken into account.

Back to Top

The Institutional and Community Environment

Finally, the environment in which the department is located also played an important role in the retention of women, as described in more detail here.

Institutional support. Statistical analyses showed that CS departments where, according to the chairperson, the department received "above average support from the institution as compared to other departments," retained female students at comparable rates to males. Because the nature of this support was unspecified in the survey, chairpersons responded to this question according to their own interpretation. Thus, the positive effect of institutional support could stem from provision of sufficient resources, including the ability to attract and retain qualified faculty, or it could stem from the chairperson's perception of a close relationship with the dean. Subsequent study of this issue should identify the specific types of institutional support that affect gendered attrition.

Local job market. The local economic environment also played a role in female retention. Departments located in regions where there were many career opportunities for CS graduates retained their female students at high rates relative to male students. The interview departments exhibited this characteristic and both correlations and regressions confirmed it. A possible reason for this positive effect is student opportunity for work experience during college.

No support for institutional selectivity as a cause of gendered attrition. In addition to all the departmental characteristics that affected gendered attrition from computer science, this study found no support for several commonly presumed factors. One of these unconfirmed factors was student quality. Reflecting Strenta's [11] finding that differences in ability and achievement were insufficient to explain low female representation in computer science, the current investigation found no evidence that CS departments in highly selective institutions had any more or less gendered attrition than departments in less selective institutions.

Back to Top

Conclusion

This investigation of Virginia's CS departments offers strong evidence that the characteristics and practices of computer science departments affect female retention at the undergraduate level. The departmental factors that affect gendered attrition are the availability of same-sex peer support; faculty characteristics and behaviors, and institutional and community environment. This study involved all 23 of the Bachelor's degree-granting coeducational computer science departments in the state of Virginia between 1992 and 1997. Restriction to one state and the small number of participating departments requires caution in interpreting and relying upon the results. A national study of CS departments will overcome these limitations. However, the Virginia study is sound and instructive, and its strengths give confidence to its findings. Chief among this study's strengths are that its conclusions were based on multiple methods of analysis, and that multiple institutions were studied. While further research is needed on a national level to corroborate this research, the quality of the current study is sufficient that its findings should not be ignored.

This article argued that inherent female characteristics are an insufficient explanation of women's underrepresentation in computer science. This conclusion has important strategic value for institutions of higher learning. Computer science departments are accountable for the retention of their female students. This accountability means that deans and chairpersons should look to their recruitment of women students and faculty; teaching and mentoring; and the context in which their CS department operates. As the demand for qualified computer professionals increases, pressure for educating women in computer science will also increase. Under these conditions, potential female students and granting agencies are likely to give more favorable consideration to departments with superior track records for retaining women throughout the undergraduate years in computer science.

Back to Top

Back to Top

Back to Top

Back to Top

Back to Top

Back to top

The Digital Library is published by the Association for Computing Machinery. Copyright © 2001 ACM, Inc.

---

No entries found