Public Management and Educational Performance:

Policies are increasingly implemented in complex networks of organizations and target populations. Effective action often requires that managers deal with arrays of other actors to procure resources, build support, coproduce results, and overcome obstacles to implementation. Few large-n studies have examined the crucial role that networks and network management can play in the execution of public policy. This study begins to fill this gap by analyzing performance over a five-year period in more than 500 U.S. school districts using a nonlinear, interactive, contingent model of management developed previously. The core idea is that management matters in policy implementation, but that its impact is often nonlinear. One way that public managers can make a difference is by leveraging resources and buffering constraints in the program context. This investigation finds empirical support for key elements of the network-management portion of the model. Implications for public management are sketched. Public Management and Educational Performance: The Impact of Managerial Networking Public policies are increasingly developed and implemented in complex networks. Analysts have documented this phenomenon in a number of countries and sketched myriad reasons for its ubiquity (Bogason and Toonen 1998; Bressers, O’Toole and Richardson 1995; Gage and Mandell 1990; Hufen and Ringeling 1990; Jordan and Schubert 1992; Klijn 1996; Marin and Mayntz 1991; Marsh and Rhodes 1992; Milward and Provan 2000; O’Toole 1998; Peterson and O’Toole 2001; Scharpf 1993). In the United States at the national level, even formal policy frequently calls for networked arrangements during execution (see Hall and O’Toole 2000). Similar trends exist at the subnational level (see Agranoff and McGuire 2000). Even in ostensibly hierarchical settings, interdependence requires that public managers deal regularly with clusters of other units to implement programs, procure resources, and gain support among stakeholders. Students of policy implementation have devoted considerable attention to the multiactor, networked aspects of implementation (see Stoker 1990; Kickert, Klijn and Koppenjan; O’Toole 2000). Ever since Pressman and Wildavsky’s classic study (1984) of economic development gone awry in Oakland, California, the challenge of the “complexity of joint action” has been prominent among serious analysts of implementation. The management dimensions of this issue, however, have been less thoroughly investigated. Many analysts recognize the importance of public managers’ mobilizing implementation action in networks (Friend, Power and Yewlett 1974; Mandell 1984; O’Toole 1983), but showing the real impact of network management is more difficult, as is determining clearly how managers operating in networks can improve implementation. Some key questions, therefore, demand attention: What difference does network-focused public management make for implementation? Do the ways managers deal with complex surroundings make a difference in how programs work? How can managers make use of their complicated and interdependent settings to enhance performance? These key issues are at the center of this article. Here we explore how public managers operate in networks, whether it makes a difference in program outcomes, and -if so how and why this might be the case. Our test case is a key policy area that has not received much recent attention in public management: public education (for earlier work see Gross, Mason, and McEachern 1958; Zeigler, Kehoe and Reisman 1985). Our approach is to investigate how managers operating in networks contribute to the educational performance of their students. The empirical setting selected for study is several hundred school districts in Texas, an especially large and diverse U.S. state. Most broadly, we are interested in the nitty-gritty ways that managing in networks might be different from managing solely in and through a hierarchy focused on program operations. Ultimately, exploring this subject can help us learn which kinds of managerial efforts might be practically useful in the complex settings increasingly a part of public administration. Public Management and Networks Public managers often operate in networked settings where program success necessitates some collaboration and perhaps coordination with other parties over whom they exercise little formal control. By network we mean a pattern of two or more units, in which not all major components are encompassed within a single hierarchical array (O’Toole 1997). Actors in networks are often located in bureaucracies that are in turn connected with other organizations outside the lines of formal authority. Many of these complex arrangements are required or strongly encouraged by policy makers, whether via interagency ties, intergovernmental links, or mandates for public-private partnership. Many others have emerged through negotiated, selforganizing initiatives of participants (Hjern and Porter 1981; Ostrom 1990). Managers may see ways to leverage their capacity for action by joining with other units in common pursuit of implementation success. Or they might conclude that others possess the financial resources or political strength to make a program more successful or more well-protected. Networks vary greatly in size, complexity, structure, and composition. They range from a simple tie between an agency and its contractor to a bewilderingly complex lattice of dozens of interlinked service providers, financing units, case-management bureaus, and support organizations -as can be found, for instance, in community mental health settings in American cities (Provan and Milward 1995). Network nodes can include public, forprofit, and nonprofit organizations, or parts of organizations. These more complicated institutional arrangements require greater managerial attention and skill, for those administering programs must link their own operations with others, tap resources in the broader network, limit potentially hostile forces, and encourage productive and collaborative partnerships. These and other network-management tasks call for actions beyond the POSDCORB injunctions of old. What do public managers in networks do and how can they make a difference? Are some managers more successful than others in such complex institutional settings? What are the managerial implications for improving implementation performance? This study focuses on a key choice made by managers: how much time and energy to work in the network, and in which directions? Managers can take a number of network-related actions, but the choice of when and where to network would seem to be especially important. Managing in Networks for U.S. Public Education U.S. education policies are implemented by locally managed school districts -usually separate, special-purpose governments without formal interdependence with other implementation units. The special-district design itself was developed precisely to buffer educational efforts from the potentially confounding actions of other governmental actors (Tyack 1974). Nonetheless, the technical and political demands placed upon school district superintendents -the chief administrators of the districts -encourage them to develop, solidify, and use ties with other important actors in their environments. The most important of these include their own school board (the elected body responsible for overall local policy), the relevant statelevel educational department (a source of funding that varies in importance from state to state, as well as the locus of many regulations), state-level legislators (who frame general education policy), local business leaders (who play crucial roles in supporting the locally enacted taxing decisions that drive much of school district revenue), and other superintendents (professional colleagues and sources of experience and innovation in the turbulent world of public education). In contemporary U.S. public education, where funding issues are critical and many supposedly separate policy problems intrude in highly visible ways in the educational process, schools have become battlegrounds for numerous policy disputes (Chubb and Moe 1990; Meier and Stewart 1991). Education reforms are frequently debated and adopted in settings where the school district is only one voice among many. Accordingly, superintendents have reason to devote managerial energy and effort to understanding and leveraging their networked environment. Superintendents manage their districts -a headquarters office along with sets of schools, which in turn are managed by school “principals” (as the term is used) -within this broader constellation of other actors, who may be potentially important as sources of funds, staff, ideas, guidance, other resources, and turbulence. The extent and kind of network to build, maintain, and use is a matter largely under control of the superintendent. Network development, then, is an opportunity available to superintendents who recognize their interdependence and try to manage it actively. The next section sketches a way to investigate whether superintendents’ directing managerial time and effort toward networking can improve implementation performance, and -more interestingly -how networking might help managers improve their implementation success. A Strategy for Exploring Network Management While extensive and growing attention has been directed to the importance of networks in public management and policy implementation, important gaps remain. Some analysts use the term “network” as a general metaphor to designate complex environments. Others describe network settings with considerable care but do not offer compelling theoretical explanations for what works and why. Still others craft vivid, instructive cases that assist our understanding of what managers do when they operate in networks and why that might be important. Even the best studies, however, have limits. They typically do not demonstrate conclusively the connection between network management, on the one hand, and implementation performance, on the other. They also are most frequently designed as case studies or, at most, a comparison of a very small number of cases. Many factors could be influencing performance in any given case, and sorting them out and checking for the distinct influence of network management per se is often not possible. Clearly needed, therefore, are large-n studies focused on network management in a way that permits controlling for other sources of program influence. In the following pages, we present a measure of network management that fits the educational program context explored in the empirical analysis, and we present evidence on its validity. Controlling for a range of other influences, we then test whether and how network management affects school districts’ performance. Then, we explore whether higher-performing organizations do things differently than those which are less impressive. And we examine patterns in how networkers deal with the opportunities and constraints in their complex settings by asking: do those more involved in the network find different ways getting things done than do other managers? To explore these questions, we use a theoretical model developed to understand public management’s impact on program performance. The model is autoregressive and nonlinear. Of special use is the model’s explicit distinction between management directed at networking in the environment of a core organization and other kinds of managerial efforts. A Theory of Network Management O'Toole and Meier (1999) provide a general model of managing programs/organizations of the following form: Ot = 1(H+M1)Ot-1 + 2(Xt/H)(M3/M4) + t [1] where O is some measure of outcome, H is a measure of hierarchy normalized to range from 0 to 1, M denotes management which can be divided into three parts M1 management's contribution to organizational stability through additions to hierarchy/structure, M3 management's efforts to exploit the environment, M4 management's effort to buffer environmental shocks, X is a vector of environmental forces, is an error term, the other subscripts denote time periods, and 1 and 2 are estimable parameters. The O'Toole and Meier model of management is autoregressive, nonlinear, and contingent. The autoregressive component is captured by the lagged dependent variable, thus requiring time series data for estimation purposes. The nonlinear elements of the model are represented by various interaction effects, some designated as reciprocal functions. The model is contingent simply because hierarchy can be considered one end of a continuum with more fluid networks on the opposite pole. As the hierarchy variable moves toward zero, the model estimates how management affects programs in network-like settings. O'Toole and Meier (1999) concede that a data set capable of operationalizing their full model does not exist and perhaps is even unlikely to exist in the future. Some theoretical gains can be made, they suggest, by testing parts of the model and adjusting those parts in response to empirical results. Our simplification of the model is twofold. We will not deal with variations in hierarchy. Although school districts vary in structure, this variation is relatively small compared to the variation that characterizes public programs in general. If we assume that structure can be treated as a constant or relatively so, the model reduces to [2]: Ot = 1(M1)Ot-1 + 2(Xt)(M3/M4) + t [2] In the model’s second term (that relating the organization to its environment), the ratio (M3/M4) is an elaboration of a general approach to managing the environment, which O’Toole and Meier term M2 in its undifferentiated form. The internal management of operations, M1, is not the prime focus of this study of network management activity, so this aspect of the full model is also excised from the simplified form we use for testing. By focusing exclusively on the general environmental management, the specific model of management considered in this paper is [3]: Ot = 1Ot-1 + 2(XtM2) + t [3] M2 is of particular interest for present purposes since it represents managerial effort in and on the networked environment.