3
Identifying Factors Thought To Contribute To Graduate Attrition
Interviews With Noncompleters
Qualitative studies have been conducted in conjunction with more formal survey work, or they may be carried out strictly in the form of case studies used initially to document relationships of interest to the investigator. Studies of attrition—whether at the graduate or undergraduate level—have frequently involved analyses of variables thought to be related to attrition. These studies often involve interviews with former students either as they exit a graduate program (Nerad and Cerny 1991) or after they have entered the workforce (Jacks et al. 1983; Arnold, Mares, and Calkins 1986; Valentine 1987) (see Text Box 6).
Longitudinal Analyses Of Linked Data Sets
A quantitative approach used with some degree of success is the analysis of degree completion patterns derived from longitudinal analyses of linked data sets. In this method, records for students in one data set may be linked with information in another data set to examine the role of "antecedent conditions" that account for differences in completion outcomes.
The methodological difficulties encountered in linking separate data sets maintained by different organizations for the purposes of longitudinal analysis are demonstrated by the paucity of studies of this kind. 11 Nonetheless, this technique has significant potential for understanding, in a systematic way, factors contributing to attrition over time.
TEXT BOX 6: THE ABCS OF ABDS In 1983 Penelope Jacks and her colleagues at the Georgia Institute of Technology published a ''narrative portrait" of 25 individuals who had completed all doctoral degree requirements except the dissertation ("all but dissertation" or ABD). Most had attended graduate school in the late 1960s, and most of those interviewed were men. Interviews were conducted by telephone with questions focused on three major issues:
The project was conducted in conjunction with a larger survey of Ph.D.s in six-fields (psychology, sociology, zoology, electrical engineering, and biochemistry) that focused on the graduate school experiences and career outcomes of scientists who actually completed their doctoral degrees around 1970 (see Porter et al. 1981). After "financial difficulties," the most frequently cited reason for leaving doctoral programs revolved around problems with advisors or doctoral committees. Some respondents reported that advisors were inaccessible owing to research work and/or travel schedules, or that dissertation committees lacked interest in a student's project. In other words, there was no one there "to encourage and give good ideas" to the students. The authors also reported that no one gave only one reason for not completing the dissertation: Often people began the interview by giving a single 'stock' explanation, but as the interview progressed several issues emerged, and the first one mentioned was not necessarily the most significant. A few respondents admitted that they had never before given a great deal of thought to the experience as a whole. The authors concluded that formal doctoral preparation may be only as effective as "the informal support system that faculty and peers provide, and in some programs, for some people, such support is never provided." Issues in Measurement The work by Jacks and her colleagues is by no means a rigorous statistical account of attrition from doctoral programs. As the authors suggest, the study may best be considered "a collective biography of would-be scientists who consented to reflect in 1980 on their experiences in graduate school." As a measurement tool, prosopography (development of the collective biography) offers the potential for augmenting statistical profiles of patterns of attrition by revealing the factors which lay behind the observed behavior. |
Educational Testing Service
A study of Ph.D. degree completion involving the linking of data sets was conducted by Rebecca Zwick at the Educational Testing Service (ETS) in 1991 (see Text Box 7). Zwick investigated the graduate careers of nearly 5,000 Ph.D.-seeking students from 11 departments in each of three major research universities. She initially invited 20 universities to determine "whether they maintained data bases that would lend themselves to the planned analyses and whether they would be willing to participate in the study." The key requirement was that information was needed at the
TEXT BOX 7: PATTERNS OF ATTAINMENT OF PH.D. CANDIDACY AND DEGREE COMPLETION In 1991 Rebecca Zwick presented an analysis of graduate school careers at three universities that examined attainment patterns of 5,000 Ph.D. students in 11 departments during the years 1978 through 1985. One goal of the study was to determine how patterns of Ph.D. attainment varied across and within the institutions. Another goal was to determine the extent to which scores on the Graduate Record Examination (GRE) and undergraduate grade point average (UGPA) predicted observed differences. A general finding was that candidacy and graduation rates in the eight years following matriculation were higher in "quantitatively oriented departments"—chemistry, physics, mathematics, and computer science—than in the humanities and social sciences. Chemistry generally showed the highest candidacy rates, while English and philosophy showed the lowest. Graduation rates revealed a similar pattern. A correlational analysis explored the association of candidacy and graduation with the GRE verbal score, the GRE quantitative score, and the UGPA. The analysis indicated that GRE scores and the UGPA were almost entirely unrelated to the achievement of candidacy and graduation. Arguing that these graduate school matriculants had already been selected on the basis of those scores, Zwick suggests: "Within the select population of graduate students, it is likely that such personality factors as perseverance … play a crucial role in determining whether candidacy and graduation are achieved." |
individual student level for at least five consecutive years of enrollment. Variables of interest to Zwick included entry date, department, citizenship, ethnicity (at least for U.S. citizens), gender, undergraduate grade point average, scores on the Graduate Record Exam (GRE), date of advancement to candidacy, and date of graduation. A further requirement was that the data be in machine-readable form. As one might imagine, attrition among the participating institutions themselves was large:
Six of the twenty schools that were contacted initially agreed to participate. In most cases, refusals were due to the unavailability of the required data. It was subsequently determined that two of the original six participating schools could not supply data on the variables of interest. A third school withdrew because no staff were available to create the needed data tapes. This left three participating schools, all of which are large research universities (Zwick 1991).
Zwick notes that for the years prior to 1982, the data base from School 1 did not include GRE scores. Therefore, with the university's permission, records from School 1 were linked to the GRE data base at ETS to obtain the scores for students in the departments selected for study. In the end, the percentage of students for whom GRE scores were available ranged from 75 to 100 percent across the 11 departments at the three schools.
The author analyzed undergraduate grade point average (UGPA) and GRE quantitative (GREQ) and GRE verbal (GREV) scores relative to admission to candidacy and graduation. Zwick reports that, in general, the prediction was "very poor," with median correlations ranging from -.09 to .11. "UGPA and GREQ were somewhat more likely to be
positively related to candidacy and graduation than was GREV." Variations across departments in the size of the correlation did not appear to follow any consistent pattern. Nevertheless, "the results do not imply that the GRE and UGPA were useless as admissions criteria" since those who were not admitted were not observed.
These studies demonstrate the utility of linking data sets in examinations of completion patterns. More formal analysis of these variables and other antecedent conditions might be considered by the research community as the opportunity to conduct this type of analysis presents itself again.
References
Arnold, Louise, Kenneth R. Mares, and E. Virginia Calkins 1986 "Exit Interviews Reveal Why Students Leave a BA-MD Degree Program Prematurely." College and University 62:34-47.
Jacks, Penelope, Daryl E. Chubin, Alan L. Porter, and Terry Connolly 1983 "The ABCs of ABDs: A Study of Incomplete Doctorates." Improving College and University Teaching 31:74-81.
Nerad, Maresi, and Joseph Cerny 1991 "From Facts to Action: Expanding the Educational Role of the Graduate Division." Communicator (May Special Edition). Washington, DC: Council of Graduate Schools.
Porter, A., D. Chubin, M. Boeckmann, T. Connolly, and P. Rossi 1981 "A Cross-Disciplinary Assessment of the Role of the Doctoral Dissertation in Career Productivity." Report to the National Science Foundation, Division of Science Resources Studies. Atlanta, GA: School of Industrial and Systems Engineering, Georgia Institute of Technology.
Tanur, Judith 1982 "Advances in Methods for Large-Scale Surveys and Experiments." In Behavioral and Social Research: A National Resource, Part II, edited by R. Adams, N. Smelser, and D. Treimon. Washington, DC: National Academy Press.
Valentine, Nancy L. 1987 "Factors Related to Attrition from Doctor of Education Programs." Paper presented at annual meeting of the Association of Institutional Research, Kansas City, Missouri, May 3-6.
Zwick, Rebecca 1991 An Analysis of Graduate School Careers in Three Universities: Differences in Attainment Patterns Across Academic Programs and Demographic Groups. Princeton, NJ: Educational Testing Service.