The analysis here was exploratory, and many limitations apply. Nevertheless, the early evidence presented here can ideally help guide further investigations of ways the university can do a better job in retaining its most valued faculty.

Colleges and universities can benefit from examining why their faculty members leave for similar jobs in other institutions. Some faculty leave for personal reasons unrelated to and unaddressable by the institution. Also, some faculty departures are desirable from an institutional standpoint and signal no underlying problems for administrative attention. Others, however, are based in significant, and perhaps avoidable, institutional problems. Unit leadership, research infrastructure, salary levels, and other factors are among the frequently cited factors in departures from academic positions (Burke, 1991). Problems in these areas, once uncovered by empirical analysis, can be addressed by thoughtful policy making and decision making, with the goal of reducing departure rates among valued faculty members. Unfortunately, knowledge about the circumstances and rationales behind the departures of valued faculty members is seldom obtained. Such information could inform efforts to improve faculty retention and productivity on a campus.

Recognizing the need for better information on faculty research productivity at the University of Minnesota, the University Senate and the Office of the Vice Provost for Faculty and Academic Personnel initiated a task force to address this concern. The task force began its work in the 2000-2001 academic year. As part of that effort, the task force requested an exploratory survey of tenure-track and tenured faculty who had voluntarily left positions at the institution for other academic positions in recent years. The request from the task force was in keeping with an earlier recommendation from the Board of Regents that the administration more aggressively examine the reasons for faculty departures from the university. The goal of the survey was to examine the factors associated with faculty leaving the university and to glean from the findings potential policy and practice implications for the university.

Below, we assay what literature there is on exit interviews and surveys in various settings, then highlight what we know about faculty satisfaction in higher education and at the University of Minnesota in particular.

The “Exit” Literature: Employee turnover is costly, and many organizations utilize exit interviews and surveys to help determine the causes and construct retention strategies for valued remaining staff. Informal contacts with top human-resources officers on campuses around the country suggest that many postsecondary institutions, especially larger research institutions, conduct exit interviews and surveys with their departing faculty.^[1] To learn more, the Office of Human Resources recently conducted an informal e-mail poll of top human-resources officials at the CIC institutions. Several CIC institutions responded. Indiana University, the University of Wisconsin, and the Ohio State University do not currently pursue exit interviews or surveys. An Ohio State official noted to us that the former effort there suffered from spotty participation and questionable representativeness. The University of Michigan does not pursue exit interviews or survey efforts at the central level but individual colleges and departments there do pursue interviews and surveys. The Pennsylvania State University and Michigan State University have long employed written exit surveys of departing tenure-system faculty who resign. Perhaps most ambitious is the University of Iowa, which sends out an exit survey (achieving a 55% return rate) and also asks emeritus faculty to conduct individual exit interviews. Thus, well-established models exist for exit interviewing and surveying in research institutions.^[2]

Perhaps understandably, however, universities do not often publicly present results from these efforts: there is very little published literature on exit interviews and surveys in higher education, and what there is tends to be largely hortatory and non-empirical. That conclusion applies to the exit literature from other organizational settings, as well.

The published literature can provide some guidance, however, on particular issues relating to exit interviews and surveys. For example, one concern immediately confronting those considering exit interviews and surveys is whether such efforts provide valid, useful information. An investigation of Fortune 500 companies over twenty years ago (Hellweg and Phillips, 1981) found that aggregate data collected in exit interviews was rarely used and little follow-up was done after termination. Companies in the sample noted the principle weakness in these interviews was the difficulty of obtaining truthful responses from exiting employees. There are parallel findings from research conducted in other corporations (Hinrichs, 1975; Garretson and Teel, 1982) and in elementary and secondary schools (Harris, 1979), although the evidence is similarly dated. Harris (1979, p. 113), for example, suggests that among departing secondary-school teachers, the expressed reason for departures is often not the reason others involved view as most “true.” Research on the validity of responses to exit interviews and surveys suffers from numerous methodological problems, however. Giacalone et al. (1997) conducted a more rigorous investigation that suggested that, while a number of factors do tend to lead respondents away from honest answers, there are ways data-gathering efforts can be designed to promote more truthful and helpful responses. Clearly, departing employees may not always want to share their full rationales, but thoughtful designs for exit interviews and surveys can help minimize dishonesty in responses.

We found no published evidence on the validity of information obtained in employee exit interviews and surveys in higher education in particular. Dworak (1983), writing about departures in higher education, notes the awkwardness of the occasion of an employee’s departure and stresses that exit interviews should address not only a need for better information but also the desirability of the exiting employee leaving with a positive impression of the organization. Information collected from departing faculty and staff can be valuable in future administrative decisions, Dworak notes, if employers take the time to analyze the data. As in other settings, however, information on departing higher-education employees may not always be used by the institutions gathering it.

With that in mind, we combed the literature for information on the relative merits of different approaches to gaining information from exiting employees. In particular, institutions must consider the relative merits of interviews (with their potential for in-depth and open-ended discussion of factors in departure) versus surveys (with their potentially lower costs, more standardized indicators, and comprehensiveness). We were curious if there was evidence on the interview/survey choice, and on the related question of whether the use of surveys may lower employees’ concerns about privacy and protection of their exiting responses. We found evidence from the corporate sector (Hinrichs, 1975) suggesting that interviews were less reliable and valid than surveys but, unfortunately, we found no evidence on these questions in higher education. Most published material has focused on exit interviews alone. We can infer, however, that surveys are especially suited for larger institutions. In campuses with only a handful of faculty leaving in any given year, survey data are likely to be skewed by individual case circumstances, unless the data are aggregated over a substantial number of years.

In all, the published research on exit interviews and surveys provided very little useful guidance for this project. This literature is relatively undeveloped, especially as it might apply to higher-education settings.^[3] With our knowledge (from the CIC “poll” and other sources) that exit interviewing and surveying are widespread in higher education and our discovery of so little information on the topic in the public domain, we concluded that what was published in analytic journals might not accurately reflect the realities of practice and policy on campuses. It is doubtful that major institutions like Iowa, Michigan State, and Pennsylvania State would continue to invest in such efforts if they were not receiving some returns from the effort. For that reason, we looked to other sources for guidance. Specifically, we sought to integrate knowledge from the faculty-satisfaction literature into our inquiry on departing faculty at the University of Minnesota.

The Literature on Faculty Satisfaction: Faculty satisfaction is often unknown until a professor chooses to leave a position in search of a better setting for his or her professional work and personal situation. Evidence from national surveys suggests, however, that faculty satisfaction is heavily influenced by certain aspects of working conditions and rewards (Fairweather, 1996; Sax et al., 1996). Faculty members’ satisfaction, and implicitly, their decisions to stay or leave, are influenced by a variety of factors. These factors may be categorized into four general domains: individual circumstances and background, departmental factors, factors at the college and university level, and factors relating to the community in which the university is located (Fairweather, 1996; Finnegan et al., 1996; Clark and Lewis, 1985).

In a 1995 national survey by the U.S. Department of Education, faculty across institutional types cited time pressures as the greatest source of stress in their working lives, followed by lack of personal time, household responsibilities, and their teaching loads (Sax et al., 1996). One area of significant potential dissatisfaction involves the possible disjunction between faculty expectations and rewards. Often, rewards and salary do not precisely align with job responsibilities. A variety of studies of faculty workloads suggest that, on average, faculty members across institutional types commit over 50 hours per week to teaching, research, service, and other responsibilities (Fairweather, 1996). In interviews and surveys, department chairs and faculty alike rate high-quality teaching strongly on their criteria for successful performance, but the reward structure sometimes sends faculty members the message that research is more valued; these disjunctions can cause stress and discontent regarding leadership and professional advancement (Clark and Lewis, 1985; Fairweather, 1996).

At the University of Minnesota, a 1997 Faculty and Staff Climate Survey (Hendel, 1999) noted that satisfaction overall was moderate among faculty: asked to respond to the statement that they were satisfied with their university employment, 62.1 percent agreed or strongly agreed, while 20.6 percent disagreed or strongly disagreed. There was substantial variation by unit in levels of agreement with the statement. 56.1 percent of faculty agreed or strongly agreed that they would accept employment at the university again. Asked about opportunities for development and advancement, faculty were largely in agreement with the statements “My current administrator/supervisor supports and encourages me in my professional development” (64.7 percent agreeing or strongly agreeing), the statement “I feel that my University supervisor/administrator has the knowledge and skills necessary to perform the job effectively (60.0 percent agreeing or strongly agreeing), and the statement “I am satisfied with my departmental/unit leadership” (53.4 percent agreeing or strongly agreeing). Most striking was the finding that 90.6 percent of faculty agreed or strongly agreed with the statement “My work gives me a feeling of personal accomplishment.”

There were areas of concern in the findings, however. Only 44.8 percent agreed or strongly agreed with the statement “I am satisfied with opportunities for advancement at the University.” While 26.2 percent expressed agreement or strong agreement with the statement that they had confidence in the university’s direction, 43.2 percent disagreed or disagreed strongly with that statement. Also, relatively few faculty expressed satisfaction with their compensation and other rewards: only 33.2 percent agreed or strongly agreed with the statement “As compared to others at this University, my compensation... is fair for the work I do” and only 18.9 percent agreed or strongly agreed with the statement “I feel that employees are adequately recognized and rewarded for their efforts and contributions here.” It should be noted that this survey was conducted during the “tenure wars,” a time of especially low morale at the University of Minnesota. Because of that timing, any generalizations from these survey results must be undertaken cautiously.

Conclusion: Although we found the literature on exit interviews and surveys scarce and not especially useful, we did find guidance in the broader literature on faculty satisfaction in higher education. The findings from our review of the literature in this area suggest that faculty members’ satisfaction and their decisions to stay or leave are influenced by individual, departmental, institutional, and community factors. We used this four-way framework as background for the analysis presented here.

Design of the Analysis

The University of Minnesota Senate and Office of the Vice Provost for Faculty and Academic Personnel jointly initiated a task force on faculty productivity at the university and subsequently provided financial support for this research project. The faculty exit survey was designed to examine sources of faculty satisfaction and dissatisfaction in the university, and suggest domains for ameliorative action. No one could argue that departed faculty are the best source of information on the factors limiting faculty productivity at this institution, but their responses in the aggregate can provide a valuable additional source of information on the question.

Instrument: On March 27, 2001, each former faculty member meeting sample-selection criteria (see below) was mailed a cover letter and a survey (see Appendices 1 and 2). The survey was constructed based on 1) the published literature on exit interviews, exit surveys, and faculty satisfaction (see above), and 2) the exit survey given departing faculty at the Pennsylvania State University.^[4] Pennsylvania State has long been committed to exit surveys of its departing faculty, and has refined its questionnaire over many years, based on the evolving uses of the data by administrative leaders there. The survey provided an extraordinarily helpful model for the survey project here at the University of Minnesota. To help build response rates, the cover letter for the Minnesota survey was sent under the signatures of both the chair of the university’s Faculty Development Task Force and the university’s Vice President for Human Resources.

The survey contained 56 items. The first 54 items were dual-response items headed “influences on faculty careers.” These items were based on 54 specific areas of influences identified earlier. These influences were broken down into four categories in keeping with the literature on faculty satisfaction (see above): individual considerations, departmental life, university and college policies and practices, and local community life. Respondents were asked to rate each influence separately for importance and satisfaction on a five-point scale, where 1 = low and 5 = high. Thus, for each of the 54 areas of influence, there were two ratings for importance and satisfaction respectively.

Two further items on the survey, termed “factors in your departure,” asked respondents to indicate the items playing the biggest factor in their decision to leave the university and whether a retention offer was made. The next item had four parts, asking respondents to provide their faculty rank, race, gender, and former university department. The final item asked for an overall rating of satisfaction with all aspects of their faculty position at the university using a seven-point scale ranging from 1 (not at all satisfied) through 4 (moderately satisfied) to 7 (very satisfied).

Sample: Sampling was limited to faculty from the University of Minnesota system’s Research I Twin Cities campus. The Twin Cities campus is one of the most comprehensive in the country with 20 colleges and professional schools, 2,275 full time tenured/tenure track faculty, and 1403 faculty on other appointment types. There is substantial variation among the four campuses in the University of Minnesota system, and creating a survey adequately addressing the missions and faculty responsibilities at all of those institutions would be difficult. Because this was an initial, exploratory study of possible future approaches to exit analyses, it was decided to begin work on a single site, the Twin Cities campus. In the report, references to “the university” are to be read as referring only to the Twin Cities campus. It is critical that similar analyses be conducted to include the other three campuses of the system.

The research sample was constructed from the records of the university’s Office of Human Resources. Tenure-track and tenured faculty who left the university’s Twin Cities campus during the preceding three years were identified as an initial pool. Then, faculty who had left the university because of tenure denial or who had left for non-academic positions and retirement were removed from the pool. Thus, the study’s core sample was tenure-track and tenured faculty who had left the university voluntarily in 1997-98, 1998-1999, or 1999-2000 for a similar position at another college or university.

College dean’s offices were asked to supply forwarding addresses for departing faculty associated with their unit in the Human Resources database. When a college could not or would not provide forwarding information,^[5] the last known address available from central records for the survey subject was used for the mailing. In total, 106 former faculty members were sent an exit survey and cover letter (see Appendices 1 and 2). Each survey was numbered, allowing the research team to send out a second letter and questionnaire to individuals who had failed to respond by April 18. By May 7, current addresses for 12 additional former faculty members had been found. These individuals were mailed the initial solicitation letter, bringing the mailed survey total to 118. At that point, we set a survey cut-off return date of June 15. By that date, 48 of the initial group of 118 former faculty had returned their surveys, constituting a response rate of 40.6 percent. Of the 48 surveys, 43 had sufficient information for sample inclusion.

Missing addresses, differential departure rates by college, and differential response rates by college all contributed to making this sample only imperfectly representative of all faculty and of all departing faculty at the university. Thus, aggregated results presented in this report should be interpreted cautiously. In Table 1, the college composition of the analysis sample is compared to that of faculty as a whole, all departed faculty, and faculty in the survey sample (i.e., departed tenure-track and tenured faculty who had available addresses and left voluntarily for other academic positions). Relative to the composition of the original survey pool, Veterinary Medicine is substantially over-represented in the analysis sample and the Medical School is substantially under-represented, because of distinctive survey response rates in the two units.

The two health units were both over-represented in their departure rates relative to their proportions of all university faculty. The Medical School, the School of Public Health, and Veterinary Medicine had especially high internal turnover rates (i.e., percentages of their faculty departing in the focal three-year period). This turnover no doubt reflects the difficult conditions of health units in this period on this and many other campuses. Relative to university faculty as a whole, Veterinary Medicine is substantially over-represented in the analysis sample, and the Institute of Technology, the College of Liberal Arts, and the Medical School are substantially under-represented.^[6]

Responding former faculty members were primarily male: 30 of the 40 who provided their gender were male. Of the 41 who responded to the item on race/ethnicity, 32 reported “white” (the nine non-white respondents included four Asians/Pacific Islanders, three African Americans/Blacks, one Native American/American Indian, and one who responded “Other.” Roughly equal numbers of respondents reporting holding university ranks at the assistant (14), associate (14), and full (15) professor ranks at the time of their departure.

Methods: Most of the 43 respondents in the final sample completed the great majority of the items on the survey. Nevertheless, only four respondents completed every item, and for two especially demanding items, only 32-35 respondents provided complete responses. This level of missing data, combined with the small overall sample, made extensive multivariate analysis (e.g., multiple regressions or factor analyses with large numbers of variables) inadvisable. Instead, simple descriptive analyses and small-scale correlational and regression analyses were conducted.

Findings
Overall, departing faculty did not indicate a high level of dissatisfaction with the University of Minnesota. The closing question of the survey asked former faculty members to indicate their level of satisfaction with all aspects of their faculty position at the university. Their mean response was 4.19 on a 7-point scale, where 1 = not at all satisfied, 4 = moderately satisfied, and 7 = very satisfied. Although only 4.7 percent of respondents stated they were not at all satisfied, 35 percent indicated a level of satisfaction below “moderate” (see Table 2). Conversely, 44.2 percent suggested they were more than moderately satisfied. One might interpret these data as evidence that a sizable number of departures are either “close calls” between faculty positions here and elsewhere or decisions rooted in factors beyond the control of the university.

There were no statistically significant subgroup differences in overall satisfaction for the departed faculty. Still, some descriptive data may be of interest. The mean level of overall satisfaction was 4.50 for former assistant professors, 3.82 for former associate professors, and 4.25 for former full professors. Recalling that these are satisfaction levels for departed professors, not in-place professors, it is striking that there are no strong differences between the most senior and most junior former faculty, and that the most dissatisfied were those in the middle rank. The mean level of overall satisfaction was 4.33 for males and 3.75 for females. Finally, the mean level of overall satisfaction for faculty of color was 4.61, compared to 4.18 for whites.^[7]

Table 2 and Figure 2 reveal that many of the departed faculty were satisfied while at the university. Table 3 suggests that, in many of these cases, the feelings were mutual. Fully 58 percent of the survey respondents indicated that before their departure, the university made them a retention offer. Of this group, 14 percent were the object of a retention offer in a previous year but not for the year of departure, 35 percent were the object of a retention offer in the departure year and declined the offer, and 9 percent were objects of retention offers in both a previous year and the year of departure. Although the proffer of a retention offer is not a perfect indicator of valuation,^[8] these results suggest that a high proportion of the respondents were valued members of the faculty.

Table 4 presents the ten factors respondents identified as the most important influences on their faculty careers at the university: (1) quality of library facilities; (2) professional autonomy (i.e., courses, research projects, and service); (3) the university’s commitment to their discipline; (4) support of department head for their continuing professional growth; (5) fairness of review processes for annual merit increases; (6) availability of time for research/scholarship, relative to other demands; (7) cultural and social opportunities in the Twin Cities community; (8) an academically strong department (re: disciplinary peers); (9) the university’s commitment to their area of research; and (10) respect and cooperation among colleagues. There was virtually no separation of the mean scores for these items, a pattern that continued deeper into the item ratings. In all, 34 of the 54 items had a mean rated importance of four or higher on the five-point scale for these items. Beyond the overall importance of so many of the items, it is intriguing to note that, of the ten highest ranked items, four came from the “university and college practices” portion of the survey, three came from the “individual considerations” portion of the survey, and three came from the “departmental life” portion of the survey.

Discerning Possible Influences on Departure – Three Approaches: Alone, neither the individual importance items nor the individual satisfaction items are fully satisfactory for providing insights into the rationales behind faculty departures. An item may be rated by a respondent as very important, but importance data without satisfaction data do not necessarily indicate a problem. Similarly, an item may be rated a source of strong dissatisfaction, but without importance data, it is impossible to discern if the factor might be a significant factor in departures. With this in mind, we developed three ways to explore the factors possibly behind faculty departures.

The importance/satisfaction gap. The first approach used the importance and satisfaction data jointly. Together, the importance and satisfaction data may help reveal the underlying dynamics in faculty departures. Because most of our respondents were patient enough to give us both kinds information for most of the items, we have a window into their reasoning in leaving the university. Accordingly, for each survey item, we constructed a new indicator subtracting the sample’s mean satisfaction rating from the sample’s mean importance rating. This new indicator , which we term the importance/satisfaction gap, is high when satisfaction levels on our five-point scale fall well short of importance levels on the same scale. Thus, items with large gaps are those for which the university experience was especially lacking for our respondents.^[9]

The data presented in Table 8 are for respondents who filled in both the importance and satisfaction items associated with a particular domain.^[10] The items with the ten largest gaps between respondents’ importance and satisfaction ratings were: (1) sense of collegiality, inclusiveness, and shared decision making; (2) the university’s commitment to their department or discipline; (3) balanced workload assignments for faculty in the department; (4) the university’s commitment to their area of research; (5) effective departmental leadership; (6) healthy social climate within the department; (7) availability of time for research/scholarship, relative to other demands; (8) effective departmental management; (9) their annual salary; and (10) fairness of review processes for annual merit increases.

Next, we analyzed overall importance and satisfaction scores, and the gaps between them, by areas of concern. That is, we collapsed the items into mean scales for each of the four areas of faculty life examined in the survey: individual considerations, departmental life, university and college policies and practices, and local community life. For consistency and conservatism, only those respondents who had no missing data on any satisfaction or importance item were included in this analysis (n=24). Table 9 presents the results. Regarding importance, the department items were ranked most important on the five-point scale, followed by community life and individual and university/college factors. Regarding satisfaction, community life was easily the most satisfying area for respondents, with the other domains a full scale point behind and clustered just above the mid-point (i.e., three) of the satisfaction scale. Translating these data into gaps brings us once again to the conclusion that it is the departmental factors that seem critical to our departed faculty. By far, the gap between importance and satisfaction was greatest for departmental factors. Well behind and closely bunched were individual factors and university and college policies and practices. Community life, in contrast, was a source of no apparent disjunction between importance and satisfaction.

To explore this pattern further, we next used multiple regression to examine further these correlational data for the satisfaction items. Specifically, we undertook stepwise regression with entry criteria set at p < .05 and exit criteria set at p < .10. This approach suffered from a high degree of multicollinearity among the independent variables as well as from the small sample and notable level of missing data.^[11] Nonetheless, the results confirmed inferences from the bivariate data. Not surprisingly, individual and departmental factors comprised six of the seven significant predictors in the final equation, with the only other factor entering being “availability of K-12 schools in the Twin Cities,” from the community portion of the survey. Overall, the final regression model explained 85 percent of the variance in overall satisfaction.^[12]

Open-ended statements of influences on departure. A third way to examine the relative weights of factors potentially influencing departure is simply to pose the question in more open-ended fashion. Question 55 on the survey asked each former faculty member to list the three most important influences on their individual decisions to leave the University of Minnesota. Faculty could use items from the survey or insert their own answers in a blank line. Four of the survey respondents chose not to answer this question, 11 respondents listed only one influence, three respondents listed two influences, 18 respondents listed three influences, and seven respondents listed four influences. Clearly, the open-ended opportunity given for this item, and the varied response possibilities, make these data imperfect, but they are instructive. Table 11 presents a list of the influences that were listed more than once. The top three influences on decisions to leave the university, as stated by individual departing faculty members, were (1) administrative problems, (2) annual salary, and (3) moving closer to family/extended family. The first and third of these were a set of open-ended responses we judged similar enough to be grouped. Other listed items follow patterns present in the earlier tables. The first and third of these items were mentioned in open-ended responses, suggesting these domains should have been more fully addressed in the formal questionnaire items. In general, responses to this item parallel the probably more reliable responses from importance and satisfaction items examined earlier, but the Question 55 data do suggest directions for future iterations of exit surveys and analyses.

The analysis here was exploratory, and many quite significant limitations apply. Nevertheless, the early evidence presented here can ideally help guide further investigations of ways the university can do a better job in retaining its most valued faculty.