(For the complete research article, click here.)
by Jason A. Grissom
Washington University in St. Louis
Superintendent turnover can hurt reform efforts and district improvement, but little research has examined what factors predict superintendent exits in a large sample. This study identifies factors contributing to superintendent exits in California by matching original superintendent and school board survey data with administrative data and information hand-collected from news sources on why superintendents left and where they went. Among 215 superintendents studied beginning in 2006, 45% turned over within 3 years. Using multinomial logistic regression to separate retirements from other types of turnover, we find that while age is the primary driver of retirements, other factors, such as school board members’ assessments of how well the board functions or whether the superintendent was hired from within the district, are highly predictive of non-retirement exits. District test score growth, however, is uncorrelated. Moreover, we find that superintendents who move migrate toward larger, higher-paying districts in more urban locations.
The story of school superintendent turnover is a well-known one: energetic new leader assumes position with plans for revitalization, only to clash with a dysfunctional school board or impatient community and move on to greener pastures before the plan can be fully carried out, leaving the district once again searching for the next great leader bearing the requisite comprehensive reform plan. High-profile examples abound of reform-minded superintendents whose tenures saw gains in student test scores but whose time in office was cut short by public
pressure and tumultuous school board relations: Arlene Ackerman in San Francisco, John Thompson in Pittsburgh, Rudy Crew in Miami-Dade County, Stan Paz in Tucson (Buchanan, 2006; Cave & Almanzar, 2008). Often, the story goes, ousted superintendents move on to other districts; Ackerman moved on to—and was then pushed out of—Philadelphia, Thompson was terminated after a year in Clayton County (Georgia) Schools, Crew had already been chancellor 2of New York City’s public school system, Paz had served in El Paso. This shuffling of superintendents through school districts creates a kind of “revolving door” in the superintendent’s office (Natkin et al., 2002), as witnessed by Kelvin Adams becoming St. Louis Public Schools’ eighth superintendent in five years in 2008 or John Covington becoming Kansas City Missouri School District’s twenty-fifth superintendent in forty years the next year (Taylor, 2008).
With chronic turnover come expectations that turnover is inevitable, making the superintendent turnover story one of short-term focus with insufficient investment in long-range vision and infrastructure (Buchanan, 2006).
The trouble with this story is that it may not be true, at least not for the typical public school district. The popular conception of the modern superintendent as a chronic mover in continual public disharmony with a conflict-ridden school board is one developed from media portrayals of prominent cases in the nation’s largest urban districts, whose experiences may not be representative of those of the suburban and rural districts that make up the majority of local school governments—or even of the average urban district. As Natkin et al. (2002) argue, this potentially errant popular understanding has consequences for both for the practice of superintendents—who become reluctant to take on major reform efforts—and the responsiveness of principals and teachers, who may adopt a “this too shall pass” approach to superintendents’ priorities and directives.
Unfortunately, there is little systematic evidence with which to question this common conception or issues of superintendent turnover more broadly, a puzzling situation given the importance ascribed to superintendents in leading district improvement. As the school district’s “chief executive,” superintendents oversee key aspects of district operations.
Research suggests that successful execution of central management functions such as staff recruitment, financial management, leadership of instruction, and strategic planning helps create positive learning environments within schools, which may indirectly impact student achievement (Alsbury, 2008; Byrd, Drews, & Johnson, 2006; Petersen, 2002). Because instability in the superintendent’s office disrupts these management functions, superintendent turnover may negatively impact district performance, at least in the short term; research concluding that successful systemic school reforms take five or more years of a superintendent’s focus suggests that negative impacts of turnover could be felt even longer (Fullan & Stiegelbauer, 1991). The loss of a superintendent may also negatively affect staff morale and satisfaction (Alsbury, 2008), which could have “trickle-down” effects on principal and teacher turnover and performance.
The importance of the district superintendent and the potential consequences of superintendent exits make understanding the factors that drive superintendent turnover a key topic for empirical research. Lamentably, however, superintendent turnover lacks a well developed research base (Natkin et al., 2002).
Existing research has primarily taken the form of qualitative explorations of turnover motivations through case studies and interviews with superintendents. Few studies have focused on empirically testing the relative strength of associations between superintendent turnover and characteristics of the superintendents, districts, and school boards with whom they work. Moreover, studies have not examined how these predictors might vary with the type of turnover (e.g., retirement, resignation).
To address these gaps in the literature, this study pulls together existing research on superintendent turnover alongside a complementary—and perhaps more well-developed—stream of research on turnover among city managers to identify potential drivers of superintendents’ decisions to leave. Research on city managers is applicable because the job of the city manager shares important characteristics (e.g., managing a complex organization, working closely with an elected board) with that of the school superintendent. We develop a simple labor market framework in which a superintendent’s continued employment in a district is determined by employment decisions made by both superintendents and their school boards, then draw on the superintendent and city manager turnover research to identify four classes of factors that contribute to those decisions: characteristics of the district, school board, and superintendent himself or herself, plus superintendent job performance.
To test the expectations our framework develops, we draw on original matched survey data from superintendents and school board members in more than 100 randomly chosen California school districts, which run the gamut from large urban bureaucracies to small rural districts with few schools, to investigate these factors in depth. These survey data are supplemented both with administrative data from the California Department of Education (CDE) and the National Center for Education Statistics’ Common Core of Data (CCD) and with additional information on superintendent turnover culled from newspapers and other public sources. Employing logistic and multinomial logit regression to model superintendents’ probabilities of turning over within a three-year window, we examine a variety of potential contributors to superintendent turnover, with attention to the differences between retirement and other types of exits.
For the complete research article, click here.
Discussion and Conclusions
This study makes several contributions.
First, it introduces the idea of considering superintendent turnover in the context of the broader superintendent labor market—in which decisions made both by school boards and superintendents are important—allowing for consideration of a broader set of factors than in most prior literature on the topic.
Second, it draws on insights from a longer public administration literature on city manager turnover to help frame turnover among superintendents, whose positions are similar to those of city managers along key dimensions.
Third, it evaluates the contributions of factors prior superintendent turnover research has suggested as important, such as the role of the school board, alongside factors that have previously gone unexamined, such as evaluations of superintendent performance.
Fourth, it separates retirements from other kinds of moves empirically to provide a more precise assessment of how different factors contribute differentially to those turnover decisions.
Finally, it sheds light on patterns in superintendents’ moves across districts, pointing toward a potentially fruitful area for further empirical inquiry.
The results paint a complex picture of superintendent turnover and one that questions some commonly held assumptions. We introduced this study with the idea that much of what we know collectively about superintendent turnover comes from cases of turnover in the largest school districts (e.g., Buchanan, 2006).
Considered alongside a representative set of school districts of other sizes, however, those districts appear atypical; three-year superintendent turnover rates in the top decile of enrollment size are approximately 30 percentage points higher than in the rest of the distribution, highlighting the need to expand our frame beyond case studies of urban districts if we wish to understand superintendent turnover more generally. Even controlling for district size and urban status, superintendent turnover is associated with other markers of a challenging district environment, such as a greater instance of student poverty. The inverse association between wealth and turnover—observed in turnover studies for other positions in school systems, including school administrators (e.g., Gates et al., 2006)—is potentially troubling, showing that turnover is higher in the districts that potentially might benefit most from stability at the top and the opportunities for sustained reform that come with it (McAdams, 1997).
Our results also question the assumption that community dissatisfaction—a dominant construct in this literature since Iannaccone and Lutz (1970)—plays a central role in most exits.
In our data, clear terminations of superintendents are very rare, though we do not discount the likelihood that some of the exits we coded as resignations were instances of superintendents being pushed out by dissatisfied school boards, which may not have been evident from the sources we consulted. Also, we do find that superintendents leave more often when the school board views their job performance less positively. Still, consistent with other studies concluding that most superintendent turnover is in fact apolitical (Alsbury, 2003), we both document the importance of retirement as a frequent source of turnover (more than a third of those we observe) and provide evidence that superintendents often move for career advancement. Despite the apparent challenges associated with positions in larger urban districts that later induce high turnover, superintendents tend to move towards positions in those districts, perhaps because such moves increase pay and prestige (Parker, 1996). Similar patterns regarding moves for professional advancement have been observed among city managers (DeHoog & Whitaker, 1990; Glass et al., 2000).
The importance of career advancement as a driver of superintendent turnover is consistent with our argument that turnover should be considered as an outcome in a labor market in which better external opportunities can attract superintendents away from their positions.
The pronounced trend for movers away from rural districts raises the question of whether such districts are used by some superintendents as “stepping stones” to more desirable positions.
Studies have documented this phenomenon among other public sector workers in rural areas, including police officers (Wood, 2001), city managers (Watson & Hassett, 2004), and school administrators (Dlugosh, 1994). Although average turnover rates are not statistically higher in rural districts, the trend we observe suggests that rural districts have less success in hiring experienced superintendents than their more urban counterparts when turnover occurs. It also illustrates that the “pull” of career advancement must be considered alongside the “push” of difficulties in the current work environment as an important contributor to superintendent exit.
Adding further complexity, evidence suggests that these two forces can operate on the same superintendent simultaneously. For example, even when a superintendent move sends him or her 36to a larger district with better pay, the superintendent often leaves behind a school board with whom the working relationship was poor (Chance & Capps, 1992).
These poor relationships with the board—pegged to board operational ineffectiveness and conflict more generally—are important predictors of superintendent exits in our study, echoing findings from prior work (e.g., Danzberger et al., 1992).
Although a key predictor, in most districts both the school board and the superintendent rate the board as high-functioning and their mutual relationship as positive in our survey data; for example, the average response to the item rating the school board-superintendent working relationship was 4.4 (on a 5-point scale) in the school board survey data and 4.5 for the superintendents (see also Glass et al., 2000).
Unfortunately, high-conflict school boards with poor superintendent relationships are more likely in (already turnover-prone) large, low-income districts (Danzberger et al., 1992; McCurdy, 1992; Grissom, 2010), compounding the likelihood that superintendents (or the board) become dissatisfied with their current employment situation and increasing the likelihood of turnover in those districts (Whitaker & DeHoog, 1991). If board dysfunction drives this sort of voluntary turnover, or if board dysfunction creates dynamics that makes them more likely to push out superintendents involuntarily, efforts to improve how well the school board works together and with the superintendent via board training or professional development may pave the way for greater leadership stability (Mountford, 2008; Grissom, 2012).
Hiring “homegrown” superintendents rather than seeking district leadership from outside may also help counteract drivers of superintendent instability. Superintendents promoted from within the district are much less likely to leave, perhaps because, as Carlson (1961) found, they are more committed to the community. Yet Carlson (1961) also found that leaders hired from inside the organization are oriented towards maintenance of the status quo and are less likely to new policies that “prepare the organization for new ways of functioning” (p. 226), which suggests that increased leadership stability can come at the expense of district reform.
Underperforming school districts may in fact be better off seeking innovative leaders from outside the district even if it means a shorter tenure is likely.
The study faces a number of important limitations. First, the data we used were drawn exclusively from California. While its diversity and size—it educates one-eighth of the nation’s students—make California a useful setting for this examination, they also set the state apart.
School districts in California tend to be more racially and economically heterogeneous and have less flexibility over resource use than districts in other states. Though the structure of superintendent and school board positions are similar to those elsewhere, we cannot be sure that our results directly generalize outside the state.
A second limitation for the study is its inability to adequately tease apart voluntary and involuntary superintendent turnover and the factors that differentially predict them. In part, this was a sample size problem; we identified too few instances of involuntary turnover to examine it separately. This inability to identify cases, however, may have reflected our data collection approach.
Involuntary superintendent turnover can be a politicized, controversial affair, and both participants and observers may have conflicting views of whether a superintendent is being pushed out or is resigning for other reasons. Formally, the superintendent may characterize a resignation as personal (e.g., to spend more time with family) or as facilitating other employment opportunities when it was, in fact, driven by board disagreement or an inevitable firing. Calls to districts, news accounts, or minutes from school board meetings may reflect only the formal story in this case, distorting our attempt to categorize.
Future studies of superintendent turnover might attempt to triangulate the reasons for turnover using a combination of approaches, including surveys of multiple stakeholders, key informant interviews, media accounts, and other sources. Such data would allow deeper examination of our conceptual model.
Future analysis in this area would also benefit from longitudinal data. The data employed here are cross-sectional, which have little power to identify causal relationships. In a cross section, we cannot be sure that the association between turnover and school board function, for example, is not driven by some unobserved characteristic of the school district (e.g., low social capital) that predicts both variables. Panel data can facilitate more rigorous analytic approaches (e.g., district fixed effects) that can help deal with such concerns. Data on superintendent employment over time would also facilitate other modeling approaches—i.e., survival analysis or a competing risks hazard model that allows for different kinds of exit—with more power to uncover the predictors of turnover at different stages in a superintendent’s tenure. Panel data can also provide means for dealing with the kinds of endogeneity that arise between turnover and variables such as salary. Though salary was a component of our conceptual model, limitations on the empirical model prevented us from treating it appropriately, though we do uncover suggestive evidence from a descriptive relationship between salary and mobility. Because salary is an obvious potential policy lever districts may pull in attempting to prevent superintendent turnover, future work that rigorously identifies the salary-turnover relationship would make a useful contribution to the literature.
Another relationship for which this study only scratched the surface was that between turnover and superintendent job performance. School board members’ subjective evaluations of the superintendent’s performance predicted turnover, but district performance did not. Perhaps superintendents are not held accountable for short-term school district test performance, which may be appropriate if, as some studies suggest, superintendents can have little effect on test scores (Ehrenberg et al., 1988). In this case, better identification of the outcome variables superintendents do affect would be a useful addition. In addition, future studies should employ strategies to account for the potential endogeneity that may arise if superintendent turnover and district performance each predict one another.
Ultimately, our analysis illustrates that superintendents exit positions for numerous reasons. As our conceptual emphasis on the two-sided nature of turnover decisions in the labor market predicts, some turnover is driven by factors that inform school boards’ decisions about future employment (e.g., superintendent performance) while other turnover comes from superintendents’ decisions to leave, which is informed by other factors (e.g., working conditions, external opportunities). Still other turnover is the result of retirement decisions, which appear to be primarily determined by age. Practically, school districts might use this results to improve superintendent retention by focusing attention or resources on its antecedents. For example, they might select candidates differently or work to build supports for superintendents faced with increased administrative complexity, such as that in unified school districts. From a research perspective, the framework and results we present point towards several new avenues for future work.