Further, the Bland et al. (2002) model suggests a hierarchical order to these three sets of qualities. That is, the individual characteristics are essential, but they have more or less power in assuring faculty research productivity depending on how research-conducive the faculty member's institution is. Finally, the impact of the institution is mediated by the qualities and style of the leader.
Many of the individual-level characteristics and institution-wide features that facilitate faculty research productivity are already present in most established research-oriented universities. For example, in such institutions research is consistently emphasized in the mission and the promotion and tenure structure. Also, most faculty in these institutions have individual characteristics, such as holding the highest terminal degree in their field, being tenured, and holding the highest rank. In addition, these faculty have most of the other individual characteristics of a productive researcher, such as being driven to do research, socialized to the research culture, and well grounded in basic content knowledge and research skills. So, although the above-cited literature is useful to institutions such as these, it is not specific enough to inform decisions about what would further facilitate the faculty's research productivity.
Statement of Purpose
The Bland et al. (2002) model addresses this inadequacy by applying a hierarchical dimension to the variables affecting faculty research productivity. Thus, using the Bland et al. (2002) model, we conducted this study to identify and confirm the broad range of characteristics associated with faculty research productivity within a single college in a research-oriented university.
As described above, the data used in this study came from a survey of 615 full-time faculty at the University of Minnesota Medical School—Twin Cities. A subcommittee of the school's faculty senate developed the questionnaire. Items were designed to assess all the characteristics in Figure 1 at the college and department levels. The questionnaire also included items on the faculty's background (e.g., degree, rank) and workload and productivity (e.g., time committed to various tasks, articles published). When possible, items were drawn from other questionnaires, such as one designed by Brocato10 while he was at the Medical College of Ohio and a questionnaire by the National Center for Education Statistics, U.S. Department of Education.11 Four specialists in faculty development were asked to assess the content validity of the items. The questionnaire was revised until it was the consensus of these experts that each item matched its intended content.
The questionnaire was then pilot tested to assure clarity and ease of completion. The final questionnaire had 56 primary questions, many with subquestions, resulting in about 150 items. With a few exceptions, the items were rated on a five-point scale with 1 = “Strongly disagree” and 5 = “Strongly agree.” For some items, respondents had the option to indicate “Don't know” or “Not applicable.” Each question corresponded to one of the three Bland et al. (2002) model clusters: individual, institutional, or leadership characteristics. Some questions, however, inherently measured more than one characteristic. For example, a question regarding the amount of time a faculty member spent on research could relate to both individual characteristics (“I choose to spend this amount of time on research”) and institutional characteristics (“The institution allows me to spend this much time on research”). These questions were ultimately linked to only one cluster by determining the primary emphasis of the question and how it related to the theoretical clusters determined in the Bland et al. (2002) model.
Survey administration, respondents, and analysis
The questionnaire was mailed in May of 2000 to all 615 full-time (those working more than 66% FTEs [full-time equivalents]) faculty, and, after several reminders, 465 were returned, resulting in a response rate of 76%. The respondents were no different from nonrespondents on demographic characteristics. On quantitative data, SPSS was used to perform t tests, logistic regressions, and multiple regressions. The University of Minnesota Institutional Review Board approved the use of these data for this study.
Results and Discussion
Of the 465 faculty who provided data, 74% were men, 45% had a high level of productivity (i.e., published five or more peer-reviewed articles in the past two years), 57% were younger than 50, and 82% held either an MD or PhD. In addition, 79% of the respondents were from clinical departments, as compared with basic sciences departments (21%); 87% were white; 99% were assistant professors or higher; and 84% were tenured or on the tenure track.
Three demographic variables in this survey were not significant predictors of faculty research productivity:
- Age. There was no difference in faculty research productivity due to age. This finding is consistent with most of the studies investigating age and research productivity that find no association.12
- Gender. Previous studies have concluded that male faculty tend to publish more than female faculty.7,13,14 At first glance, we found male faculty published more than females, but this difference was eliminated when the density of female faculty in lower ranks was taken into account. Thus, in the present study, we found no difference in productivity due to gender when rank is controlled.
- Department type. Additionally, we found no difference in productivity due to department type (clinical versus basic science). Notably, the clinical departments have both MD and PhD faculty. The most research-productive faculty subgroup in this sample was PhDs in clinical departments.
Conversely, we found two demographic variables were significant predictors of research productivity:
- Appointment type. We found a significant difference in faculty research productivity according to appointment type (tenure-track faculty were more research productive than were faculty on other appointments). This is consistent with previous studies.15
- Rank. We found a significant difference in research productivity by rank (e.g., full professor, associate professor, and assistant professor). Because research productivity is one of the major criteria for promotion, high research productivity in faculty of higher rank is understandable. Again, this finding is consistent with previous studies.7,16
Predictors of faculty research productivity
What individual, institutional, and leadership variables best predict a faculty member's research productivity?
We conducted a multiple logistic regression to summarize the relationship between a faculty member's level of research productivity and individual, institutional, and leadership qualities (see Table 1). The outcome variable was faculty who produced more than five articles in the last two years, called highly research productive, versus those who produced fewer, called moderately research productive. Demographic variables, except rank, were included in the regression (e.g., gender, age, time spent on various activities, degree, race). The regression found that data on nine questions resulted in correctly assigning 75% of the respondents to the categories of highly or moderately research productive.
Six of the nine variables in Table 1 were positively associated with research productivity, meaning that high faculty responses to these questions were associated with higher research productivity. The six variables were:
- I would describe myself as being internally driven to conduct research.
- On average, in 1999, how many hours each week were you involved in research?
- I have a well-developed network of colleagues with whom I discuss research projects and education.
- On average, in 1999, how many hours each week were you involved in administration?
- I have been (or was when I was a junior faculty member) formally assigned an advisor or mentor within my academic department.
- The number of faculty in my department is large enough to accomplish our goals in research.
Most of these variables intuitively make sense. The variable regarding spending a larger amount of time on administration may have emerged in the model because 70% of our department heads fall into the category of highly research productive. These faculty members, all strong researchers, would also have listed large numbers of hours in administration. Also, our highly productive researchers often manage large grants, requiring them to spend time on such things as personnel, budget, and other administrative duties.
Three variables were negatively associated with research productivity, meaning that low faculty responses to these questions were associated with a high level of research productivity. These variables were:
- A large portion of my academic department's faculty can be considered to be a significant external grant-getter.
- On average, in 1999, how many hours each week were you involved in teaching?
- I have a well-developed network of colleagues in the department with whom I discuss research projects and education.
These results suggest that having a network within the department is not necessary for research productivity, whereas having an external network is essential. This external network of highly productive researchers likely becomes the frame of reference for a highly productive researcher. Thus, it is possible that researchers then view their own department in comparison and see that it does not possess as large a number of significant grant-getters as does their external network. Having fewer hours in teaching reflects the reality of highly productive researchers committing more time to research compared with others.
Of course, we do not know if these variables “cause” high research productivity; we only know they are associated with high research productivity. However, the findings do suggest that these individual variables are ones to consider when trying to create and implement faculty development programs and other types of initiatives to facilitate research productivity.
What individual, institutional, and leadership variables best predict group (department) research productivity?
Faculty do not work in isolation. The primary environment in which they work is their department, and when research is the goal, they are advantaged to be in a department that is collectively research productive.7,17–21 To identify differences between highly research productive departments and moderately research productive departments, we ranked departments by their percentages of faculty considered to be highly research productive. Data analysis revealed a natural break in the data between 42% and 48% of the department's faculty; that is, there were no departments where 43–47% of the faculty were highly research productive. Thus, we defined departments as highly research productive if 48% or more of their faculty had published five or more articles in the last two years. Furthermore, we defined departments as moderately research productive if 42% or less of their faculty had published less than five articles in the past two years.
Interestingly, we found 11 of 23 departments (48%) were highly research productive; 56% of our faculty were in these departments. In fact, all but two of our large departments (i.e., 20 full-time faculty or more) were in the highly research productive group. Previous research found a positive association between size of department and level of department research productivity.8,19,21–24 In our study there is a small, nonsignificant relationship between department size and productivity (r = .26).
We performed a multiple logistic regression using all the questionnaire's items related to the department, with the dependent variable being moderate versus highly research productive department (see Table 2). Again, all demographic variables except rank were also included in the regression (e.g., gender, age, time spent on various activities, race). The regression model (Table 2) correctly assigned a faculty member 78% of the time to the correct level of department productivity.
Thus, we found a department with high research productivity in our school was one where the core missions of research and education were emphasized, the expectations to be productive in research and generate external research dollars was clear, a large portion of the faculty were grant-getters, there was good communication about issues, and there were nonmonetary mechanisms for rewarding teaching. At the same time, we found a negative association with having a department head known best for his or her administration, where recognition and nonmonetary rewards were provided for administration, opinions were seriously considered by leadership when making important decisions, and more time was spent on teaching.
This regression revealed the important role of the department head. He or she keeps the core missions in front of faculty, makes the generation of dollars through research a high expectation, and assures communication. Similarly, department heads have a great deal to do with the number of hours faculty devote to teaching as well as other roles and what work is recognized nonmonetarily.
What individual, institutional, and leadership variables best predict a faculty member's satisfaction?
To maintain high research productivity, an institution presumably must retain productive researchers and attract promising new faculty. To summarize the relationship between research-conducive characteristics and faculty members’ satisfaction with their department, we performed a stepwise multiple regression analysis using as the outcome measure the response to the question: “If I were to select a faculty career again, I would choose to be: in my current department.” Fifteen variables accounted for 53% of the variance in faculty's satisfaction with their department and are listed in Table 3.
Satisfaction with a department seems to be primarily associated with institutional and leadership variables rather than with an individual faculty's variables. Perceived support from the department head for both faculty's teaching and research efforts is particularly important, as is having one's opinions seriously considered and having opportunities to pursue research interests. Having capable colleagues is also important, as is reflected in having recruitment strategies to attract talent, having a sufficient number of faculty, and having a network of faculty.
When a department expects faculty to generate as much revenue as possible from nonteaching or nonresearch activities such as patient care, or faculty to spend more hours on education and research, faculty satisfaction is significantly impacted. The negative correlation with more research hours may be a result of responses from basic scientists who, on average in this study, were less satisfied than were clinical faculty and who spent more time on research (32 versus 18 hours per week).
How well does the Bland et al. (2002) model explain research productivity?
The Bland et al. (2002) model is designed to explain faculty research productivity on two levels. First, it suggests there are specific individual, institutional, and leadership characteristics associated with faculty research productivity. Second, it suggests there is a hierarchical order to these three groups of behaviors. That is, the individual characteristics are essential, but they have more or less power in assuring research productivity depending on how research conducive the institution is. Finally, the impact of the institution is mediated by the qualities and style of the leader.
The regression that predicted an individual's research productivity (see Table 1) contains one of the individual characteristics (motivation) and eight of the institutional characteristics in the Bland et al. (2002) model. It is not surprising that the regression did not include additional individual characteristics, as has been the case with many other research studies on faculty research productivity. Recall that nearly all the faculty in this study possessed the individual characteristics associated with research productivity, such as content knowledge, advanced research skills, and scholarly work habits. Thus, there was little variance on these characteristics. Being driven to conduct research, labeled “motivation” in the model, was also consistently high for all faculty studied (the modal response was four on a five-point scale where five was high). However, the variance on this item was striking: nearly all the highly productive faculty rated this item five. In short, they were very driven to conduct research.
The other eight items that predicted individual research productivity picked up many of the institutional characteristics in the Bland et al. (2002) model, such as faculty having a formally assigned mentor, a well-developed network of colleagues outside the department with whom to discuss research and education, time to do research, less time on teaching, enough faculty to achieve goals, and a less-developed network of department colleagues.
Leadership characteristics were not identified by this regression; however, it is unlikely that the institutional characteristics would be present in the regression without a leader who possessed the characteristics in the model.
If a model for faculty research productivity was based solely on this regression, it would use the characteristics noted with a superscript of “1” in Figure 2. But, a model using only these characteristics would not be very useful and would likely be incomplete because most researchers conduct their work as members of a larger organization. Further, most of the predictors in Figure 2 are not controlled by the individual but rather are a function of the organization in which they work. In addition, these features are highly dependent on the group in which the researcher works; for example, are there enough faculty to do the work and an adequate number of senior faculty who can be mentors? For that reason, to understand what facilitates individual research productivity one also needs to look at features of the department that facilitate group productivity.
Our second regression (see Table 2) did just that. This regression picks up many of the institutional and leadership characteristics found in the Bland et al. (2002) model. If a model were built on just this regression, it would include the characteristics noted with a superscript of “2” in Figure 2.
Together, these two regressions and two models confirm a large number of the characteristics suggested in the Bland et al. (2002) model (see Figure 2). Recall that the Bland et al. (2002) model was developed via a synthesis of studies on research productivity that varied widely in their study designs and populations. It is reassuring that so many of the features drawn from such a wide range of studies are confirmed as essential even in a very specific type of institution (i.e., a medical school in a research-oriented university).
We conducted a confirmatory factor analysis to affirm that the variables we observed do cluster into our three theoretical groupings. Thus, for this analysis, we used only the questions related to research. The overall fit of the model was good. All of our variables loaded significantly on their theoretical factors, with two exceptions. Significant factor loadings ranged from .14 (“mentoring,” a dichotomous variable) to .79. Bentler's Comparative Fit Index was .96, and the more conservative Normed Fit Index was .91 (for which, in general, any amount over .90 is considered good).25,26 The root mean square error of approximation (RMSEA) was .056 (an RMSEA under .06 is considered good).27
As mentioned, two characteristics—having a well-developed external network and number of hours spent on research—did not load significantly on the institutional factor, as we had expected. Rather, these loaded significantly on the individual factor (Table 4 displays the loadings on both the individual and institutional factors). In the evolution of the Bland et al. (2002) model, it has been noted that some of the characteristics associated with research productivity would actually be linked to two different groupings. For example, “hours spent on research” is likely a function of both the institution and individual and would, therefore, fit in both factors. Thus, as the model has evolved, some characteristics have moved between groupings. In addition, we found that two other items loaded significantly on two factors (the predicted factor and one other). “Department commitment,” and “communication with colleagues” loaded on both the individual and institutional groupings.
These empirical results suggest that “external network” and “number of hours spent on research” should be moved in the model to the individual factor and that “department commitment” and “conversations with colleagues” should be placed in two groupings, although there is an attractive simplicity to placing each characteristic in its dominant cluster. Before moving these characteristics, it is important to consider the context of this study. At the University of Minnesota Medical School—Twin Cities, there is no written or unwritten policy as to the amount of time faculty are expected to devote to research. Rather, each faculty member negotiates research time with the person to whom they report, and the agreed upon time is largely dependent on the amount of external research dollars the faculty member has acquired. Similarly, opportunities to network with colleagues are largely a function of the travel dollars faculty members have as a result of grants. Thus, at our school, variations in time spent on research and networking with external colleagues are largely determined at the individual level, not the institutional level. However, many schools have at least a minimum percentage of time that all faculty are expected to devote to research and a minimum amount of travel dollars for each faculty to attend professional conferences. If that is the case, the institution creates the opportunity for networking with colleagues and determines the amount of time faculty can spend on research. Had our survey included faculty from different schools, we believe these two characteristics would have loaded on the institutional grouping. Therefore, we are not convinced that the Bland et al. (2002) model should be modified based on the results of this one study.
Further, the Bland et al. (2002) model suggests the characteristics associated with research productivity cluster into three unique groupings: individual, leadership, and institutional. To assess this, we looked at the correlations between them. We found that the correlation of the individual characteristics with leadership characteristics was r = .21; the correlation of individual characteristics with institutional characteristics was r = .39; and the correlation of leadership characteristics with institutional characteristics was r = .95. So, the clusters appear unique except for the last correlation of institutional and leadership characteristics. However, this finding is likely an artifact of the study data (the respondents were from the same institution and school). There could be no institutional variability except on the departmental level. Thus, in this study, the leadership variable was highly constrained. Because leadership at the dean's level and above was fixed, we could reflect only upon leadership differences at the departmental level and below.
To confirm that institutional characteristics and leadership characteristics are truly different, even under these conditions, we ran a second factor analysis combining the leadership and institutional factors. This second analysis, the two-factor model being a subset of the first (the three-factor model), did not fit the data as well (χ2(2) = 133.19; p < .001). This confirmed that the leadership and institutional factors are statistically different, even though they are highly correlated in this situation. And, it confirmed that the three-factor model is the best fit with the data.
Finally, we turn to the hierarchical order of these three groups suggested by the Bland et al. (2002) model. Unfortunately, the productivity measure in this study is a simple dichotomous variable and is inadequate for a path analysis to assess, statistically, the hierarchy of these groupings. So, with these data, we are unable to either confirm or disprove the hierarchical assertion in the Bland et al. (2002) model.
How is this information practically applied in facilitating individual or group research productivity?
To use this information to facilitate research for one's self, for another individual faculty member, or for a group or department, one needs to combine the findings. An individual cannot be productive in isolation. On the other hand, one cannot have a research-productive group without having faculty members who are individually productive. Thus, we return to the Bland et al. (2002) model. It accurately contains the characteristics associated with research productivity. However, the lines showing the characteristics’ direction of influence do not completely reflect the regression results (see Figure 1). For example, we did not confirm that motivation for research, an individual characteristic, affects research productivity only if it is coupled with essential institutional characteristics. Yet, we know that motivation cannot result in research productivity without the many supportive characteristics found in the institution. The Bland et al. (2002) model integrates the statistical findings from the two regressions with our conceptual understandings about research to provide an integrated picture that can guide those wishing to increase faculty research productivity.
As previously mentioned, a significant amount of disparate research has been conducted on variables, especially at the individual faculty level, to determine what leads to research productivity. Drawing from this literature, Bland et al. (2002) proposed a model that identified a number of important factors at the individual, institutional, and leadership levels. We sought to confirm how this broad set of individual, institutional, and leadership variables operate together to impact faculty research productivity within the context of a single research university.
The separate analyses of the characteristics associated with individual research productivity, group productivity, and faculty satisfaction not only identified the characteristics that predict these outcomes, but also demonstrated how separate these outcomes are. Quite different characteristics are associated with each of these outcomes.
With regard to individual faculty's research productivity, the predicting characteristics are very similar to what other studies have found. In some ways this is surprising, given that this study was done in a highly research-oriented institution with quite established faculty. However, it seems that even in this type of institution, when individual faculty's research productivity is the goal, nothing substitutes for recruiting faculty with a passion for research, providing them with formal mentoring programs, facilitating their networks, and providing time for them to do research. It also confirms that an individual's research productivity is influenced by a combination of individual characteristics and institutional characteristics. We would also contend that this confirms the importance of research-oriented leaders. Even though leadership factors did not specifically load, the institutional features that did are primarily in the hands of administrators. Also, the leadership features in the model are highly correlated with the institutional features.
The characteristics that facilitate group productivity closely mirror the institutional and leadership characteristics in the model. Although individual characteristics did not load on this regression, this does not suggest that these are unimportant to departmental research productivity. In fact, individual characteristics are the foundation of and prerequisite for many of the institutional variables including culture; positive group climate; size, experience, and expertise; and mentoring. Nevertheless, institutions that want most of their faculty, instead of a few stars, to be highly research productive should emphasize institutional and leadership characteristics such as clear coordinating goals, research emphasis, communication, and assertive–participative governance.
Taken together, these separate analyses reinforce the perception that a highly research productive organization is indeed a function of the integration and interplay of the individual and institutional features. Furthermore, the successful synthesis of these features is heavily dependent on effective leaders. It is this combination that the Bland et al. (2002) model depicts.
1 Bland CJ, Seaquist E, Pacala JT, Center B, Finstad D. One school's strategy to assess and improve the vitality of its faculty. Acad Med. 2002;77:368–76.
2 Bland CJ, Schmitz CC. Characteristics of the successful researcher and implications for faculty development. J Med Educ. 1986;61:22–31.
3 Bland CJ, Ruffin MT, IV. Characteristics of a productive research environment: literature review. Acad Med. 1992;67:385–97.
4 Bland CJ, Bergquist WH. The Vitality Of Senior Faculty Members: Snow On The Roof–Fire In The Furnace. ASHE-ERIC Higher Education Report, 25(7). Washington, DC: George Washington University, Graduate School of Education and Human Development, 1997.
5 Bland CJ, Weber-Main AM, Finstad D, Lund SM. Research Productive Departments: Strategies from Departments that Excel. Bolton, MA: Anker Publishing Co., 2005.
6 Finkelstein MJ. The American Academic Profession: A Synthesis of Social Scientific Inquiry since World War II. Columbus: Ohio State University Press, 1984.
7 Creswell JW. Faculty Research Performance: Lessons from the Sciences and Social Sciences. Washington, DC: Association for the Study of Higher Education, 1985.
8 Dundar H, Lewis DR. Determinants of research productivity in higher education. Res Higher Educ. 1998;39:607–31.
9 Teodorescu D. Correlates of faculty publication productivity: a cross-national analysis. Higher Educ. 2000;39:201–22.
10 Brocato JJ. The research productivity of family medicine department faculty: a national study [dissertation]. Michigan State University, 2001.
11 Russell SH, Cox RC, Boismier JM. A Descriptive Report of Academic Departments in Higher Education Institutions. 1988 National Survey Of Postsecondary Faculty (NSOPF-88). Washington, DC: U.S. Department of Education, Office of Educational Research and Improvement, 1990.
12 Blackburn RT, Lawrence JH. Aging and the quality of faculty job performance. Rev Educ Res. 1986;56:265–90.
13 Cole JR, Zuckerman H. Marriage, motherhood, and research performance in science. In: Zuckerman H, Cole JR, Bruer JT (eds). Outer Circle: Women in the Scientific Community. New York: W. W. Norton & Company, 1991:157–70.
14 Sax LJ, Hagedorn LS, Arredondo M, DiCrisi FA III. Faculty research productivity: exploring the role of gender and family-related factors. Res Higher Educ. 2002;43:423–46.
15 Bland CJ, Finstad D, Center B, Staples J. The impact of appointment type and faculty role differentiation on the productivity, satisfaction, and commitment of full-time faculty in research/doctoral institutions. J Higher Educ. [submitted].
16 Creswell JW, Bean JP. Research output, socialization, and the Biglan model. Res Higher Educ. 1996;15:69–89.
17 Bland CJ, Schmitz CC. An overview of research on faculty and institutional vitality. In: Schuster JH, Wheeler DW (eds). Enhancing Faculty Careers: Strategies for Development and Renewal. San Francisco: Jossey-Bass, 1990:41–61.
18 Pelz DC, Andrews FM. Scientists in Organizations: Productive Climates for Research and Development. New York: John Wiley & Sons, 1966.
19 Pineau C, Levy-Leboyer C. Managerial and organizational determinants of efficiency in biomedical research teams. In: Epton SR, Payne RL, Pearson AW (eds). Managing Interdisciplinary Research. New York: John Wiley & Sons, 1983:141–63.
20 Andrews FM (ed). Scientific Productivity: The Effectiveness of Research Groups in Six Countries. Cambridge: Cambridge University Press, 1979.
21 Blackburn RT, Behymer CE, Hall DE. Research note: correlates of faculty publications. Sociol Educ. 1978;51:132–41.
22 Baird L. What characterizes a productive research department? Res Higher Educ. 1986;25:211–25.
23 Johnston R. Effects of resource concentration on research performance. Higher Educ. 1994;28:25–37.
24 Jordan JM, Meador M, Walters SK. Effects of department size and organization on the research productivity of academic economists. Econ Educ Rev. 1988;7:251–5.
25 Bentler PM, Bonnett DG. Significance tests and goodness-of-fit in the analysis of covariance structures. Psychol Bull. 1980;88:588–606.
26 Garson GD. Structural equation modeling 〈http://www2.chass.ncsu.edu/garson/pa765
〉. Accessed 12 February 2003. PA 765, Statnotes: An Online Textbook. College of Humanities and Social Sciences, North Carolina State University, Raleigh, NC.
© 2005 Association of American Medical Colleges
27 Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling. 1999;6:1–55.