Public Management on the Ground: Clustering Managers Based on Their Behavior

ABSTRACT

Public management research has identified a dizzying array of management variables that affect organizational performance. While scholars have learned much by analyzing one or a few specific behavioral dimensions of public management at a time, we argue for the value of a more holistic and inductive approach that uses data on several aspects of public management for identifying manager types. Such an approach accounts for both the cognitive processes of people affected by management and the reality that managers’ individual behavioral decisions are interrelated. We examine the overlap of 21 aspects of public school management behavior using cluster analysis. We identify four different manager types (“firefighters,” “laissez-faire managers,” “administrators,” and “proactive floor managers”), each reflecting a distinct constellation of managerial behaviors. The manager types we call “administrators” and “proactive floor managers” are associated with relatively better outcomes, while “firefighters” are associated with relatively worse outcomes.

Notes

Cluster analytical findings are sensitive to the variables included. Adding a new variable, such as another management variable, might make respondents at the edge of one cluster move to another. For this reason, we do not use (and do not suggest that others use) cluster analysis to make claims about the frequency with which managers adhere to the practices of particular clusters. The main value of cluster analysis relates to its ability to identify manager types, each marked by distinct constellations of managerial behaviors. We acknowledge the limitations of cluster analysis, but we argue that the method is useful for taking a holistic approach to studying public management that accounts for the complexity and interrelatedness of managerial behaviors.

The students’ achievements in Danish are measured by the mean product of three test scores (reading, writing, and spelling); their math achievements by the mean of two test scores (arithmetic and mathematical problem solving). The subjects are given equal weight in the final student performance measure.

As with the SFI school principal survey, all public lower secondary schools in Denmark were surveyed (1,478). The response rate was 52%, for a total of 767 respondents.

A total of 1,998 teachers teaching one or more ninth-grade classes in Danish or math in the school year of 2010–11 were surveyed. The response rate was 57%, yielding 1,130 teacher respondents.

As we are not interested in the hierarchical relations of clusters, we use kmeans as opposed to hierarchical clustering methods; e.g., single-linkage, average-linkage, complete-linkage, Ward’s method.

We ran four kmeans cluster analyses, respectively specifying k at two, three, four, and five. For each cluster solution, a Calinski-Harabasz pseudo-F index score was computed (with larger values indicating a more empirically distinct cluster structure). Increasing k was associated with decreasing index scores. However, the score differences were relatively small. Given the relevance of a clustering resolution exceeding a binary distinction, we thus decided on a four-group solution similar to existing typologies (e.g., Miles and Snow 1978).

Analysis of the pairwise correlations among the four outcome measures reveals the following: teacher absenteeism is negatively related to student performance (−.20, p < .1), while teacher goal commitment and job satisfaction are positively associated with one another (.33, p < .001).

We conducted a robustness test in which we excluded managers at edges of the clusters from the sample. To identify these managers, we estimated the (Euclidean) distance between each manager’s responses and the mean responses for the manager’s own cluster. We then dropped the five percent of managers with the largest distances (between themselves and their own clusters’ means) and re-ran the regression models. The results are very similar to those reported in Table 3. The robustness test thus indicates that our findings are not driven by managers who do not fit particularly well into any one cluster.

Additional information

Notes on contributors

Mogens Jin Pedersen

Mogens Jin Pedersen ([email protected]) is a senior researcher at VIVE – The Danish Centre of Applied Social Science and a postdoctoral researcher at Department of Political Science, Aarhus University. His research focuses on work motivation, public management and performance, employee behavior and decision making, gender and racial biases, and research methodology.

Nathan Favero

Nathan Favero ([email protected]) is an assistant professor in the Department of Public Administration and Policy at American University’s School of Public Affairs. His research interests include public management, race and ethnicity, public policy, public administration, and research methodology.

Vibeke Lehmann Nielsen

Vibeke Lehmann Nielsen ([email protected]) is a professor at the Department of Political Science, Aarhus University. Her research interests include gender in public organizations, public employee behavior, public manangement, and citizen behavior in interaction with public authorities.

Kenneth J. Meier

Kenneth J. Meier ([email protected]) is a distinguished professor of public administration at Texas A&M University and professor of public management at the Cardiff School of Business (Wales). His current research interests focus on race and gender in public organizations, testing the generality of public management models across national contexts, and sector differences in the link between management and performance.