A WORLD OF NETWORKS—AND SOME CAVEATS

So I'll refer several times to some of the results of that research program but turn instead to a related theme, one that has drawn my research attention in one way or another for more than 25 years: the networked character of public administrative action and the externally oriented networking behavior of the managers who try to move such complex arrangements toward productive action. Networks and networking are all the rage these days (Borgatti and Foster Reference Borgatti and Foster2003)—in some respects, probably too much so. (Check out the Web site and letterhead of the American Political Science Association: the emblazoned slogan is “Networking a World of Scholars.” And at JPART, the most highly rated of the public administration journals, a new position of “co-editor for networks” has just been created, with Keith Provan recently appointed.) But I want to argue here that research in public administration needs to pay systematic attention to the implications of networked action for traditional concerns of the field—and also that such networks and networking carry implications for a number of key issues in political science research. So public administration and political science are bound partially together by forces and developments of mutual influence, even if we seldom attend to that reality.

Networks of (usually organizational) actors are all around us in the world of public administration, whether for treating the seriously mentally ill (Provan and Milward Reference Provan and Milward1995), or executing programs for family planning (McFarlane and Meier Reference McFarlane and Meier2001), or dealing with the homeless among us (as with the National Alliance to End Homelessness), or trying to manage important aspects of the national and international systems of finance (Cassell and Hoffmann Reference Cassell and Hoffmann2009). John Gaus himself, writing before I was born, noted that “city people little appreciate the great network of agencies for agricultural research and for making the new knowledge available for farmers” and others (Gaus Reference Gaus1947, 27).

So let's begin with some basics. Networks are structures of interdependence involving multiple organizations or parts thereof, where one unit is not merely the formal subordinate of the others in some larger hierarchical arrangement (O'Toole Reference O'Toole1997, 45). And externally oriented networking behavior on the part of public managers can perform a number of functions—like generating support, negotiating with other actors in the networked environment, acquiring information, tapping opportunities, defending the core organization from threats, and sometimes helping to move a network of interdependent organizations toward an objective.

Networks typically do not replace bureaucratic organization; but they add a layer of sometimes not-so-visible structural complexity, as government agencies are interwoven with counterparts from the same government, or other governments of the same sort—as with multiple governments in metropolitan regions (Frederickson Reference Frederickson1999; Feiock and Scholz Reference Feiock and Scholz2010)—or from other levels of governments, or even from ministries and agencies that are a part of national regimes in other countries (Nicolaidis and Howse Reference Nicolaidis and Howse2001). They may also be linked via contracts or other ties with for-profit or nonprofit organizations. Or even all of the above.

So individuals as actors can be seen simultaneously as occupants of positions within a public administrative organization and also as one component of a multi-organizational web of action built in one way or another around a function or public problem.

It has sometimes been claimed that network researchers assume that this structural form is supplanting hierarchical organization (for instance Olsen Reference Olsen2004), but this is untrue. The observable dynamic is rather more interesting: public administration now confronts the challenge of coordinating action both within and across organizational boundaries, with the use of formal authority and also via other means (Hjern and Porter Reference Hjern and Porter1981).

Commentators sometimes jump to the conclusion that networks have appeared on the scene very recently—that such arrays are a novel and emergent structural form. But this claim is contestable, as the quoted comments from John Gaus suggest. Thad Hall and I have demonstrated, for instance, that U.S. federal statutes establishing new public programs or substantially revising existing programs require or strongly encourage multi-organizational action most of the time—85% of the cases in our sample fall into this category. More relevant for present purposes, the extent of such networked action, as measured by statutory and regulatory content, was basically the same for two Congresses 28 years apart (Hall and O'Toole Reference Hall and O'Toole2000, Reference Hall and O'Toole2004). So it may well be that we have been fairly blind to these aspects of administrative structure and behavior (and their influence) until recent years.

Three more points of possible misunderstanding to get out of the way:

First, networked forms of public action are not necessarily voluntary—that is, self-organizing—nor even necessarily cooperative. The point about mandated networked action is effectively made by the findings, just mentioned, regarding statutory directives or encouragements. A role of governments has clearly been to force, or at least press toward, networked forms of administrative action—sometimes with direct hierarchical control or regulation as a default, upon failure of cooperative effort. And the point about cooperation can be made in many ways. In game-theoretic terms, it is clear that between or among interdependent actors, pure cooperation and pure conflict are extreme forms. The vast majority of possible “ties that bind” involve mixed motives, with those involved neither in full agreement nor in strident opposition, but somewhere in between. So too it is with organizational actors in networks (see Stoker Reference Stoker1991). Except for special cases, the pattern involves sets of partially overlapping and partially distinct goal schedules—not to mention world views, jargon, funding constraints, staffing patterns, missions and mandates, and so on. Indeed, they may be tied together in “linked games” (Tsebelis Reference Tsebelis1990) across several different programs and with other sets of additional actors involved, depending on the program.

While some forms of network-based collaborative action may take on characteristics approaching voluntaristic, self-organized, and highly consensual joint effort (Ostrom Reference Ostrom1990), most network patterns of interest to public administration are more complex. They may contain elements of both mandated and self-organized ties between organizations; they may contain public, private, and nonprofit units; and they typically include actors with somewhat different objectives—or at minimum objectives with different levels of salience across the actors involved. And, as Robert Montjoy and I observed some time ago, they may be linked by somewhat different kinds of ties—at the most abstract level, these are authority, common interest, and exchange (O'Toole and Montjoy Reference O'Toole and Montjoy1984). The combination of complex patterns that are partially mandated and partially self-organized suggests an intriguingly complicated terrain, and one that cannot easily be navigated, theoretically speaking, with many of the approaches often used.

Second, critics of a network perspective on public action sometimes suggest that network researchers carry an overly benign view—that network analysts root for collaboration and sharing and neglect the hard-headed realities of what else can and does happen in and around networks. I want to argue that it is important to encourage research from a network perspective, but saying so does not necessarily entail any endorsement of network forms or any carte blanche for whatever happens in and through them. Indeed, as I argue here, networks and networking can encourage and produce beneficent outcomes—and also worrisome ones. Doing network research does not entail being an ideologue or a naive social observer. One straightforward way of seeing that networks can produce both benign and malign results is to consider what is meant by expanding public action from lone organizations to patterns of interdependent units. Abstractly speaking, adding such actors does multiple things: it enhances capacity and also complicates the set of goals—meant as constraints on action (Simon Reference Simon1964); further, it can render less visible just who is doing what to whom. Whether one of these tendencies outweighs the others in a given case is dependent on the details. So we can be very interested in networks and public administration without being cheerleaders for unfettered cooperative public action.

Third, emphasizing the relevance of networks and managerial efforts outward toward the interdependent environment of public organizations does not mean that the traditional attention to internal management is passé. In fact, management within such organizations remains important, and we have been able to demonstrate that traditional internal functions represented in the “M ₁” term of our formal model—like managing the organization's human resources and budget—can be shown to contribute in important ways to organizational performance (O'Toole and Meier Reference O'Toole and Meier2009; Meier and O'Toole, forthcoming). We shouldn't make the mistake of seeing the question in either/or terms, since doing so systematically under-specifies public managerial effort and influence. Managing organizations matters—and so do networks and external managerial effort in the networked environment.

Some time ago, I sketched a number of research agendas that “treating networks seriously” should include (O'Toole Reference O'Toole1997). I shall not cover that ground again here. What I can say is that in the years since that argument appeared in print, considerable research progress has been made on at least some of the key questions. We are beginning to have substantially more systematic descriptive information about the presence and shape of networks as they pertain to public administration—as well as their involvement in policymaking and other political processes. We have data, for instance, on public program and public-sector-focused network forms in U.S. local, state, and national governance and British, Dutch, Swedish, Finnish, and Thai administrative settings, among others (respectively, Agranoff and McGuire Reference Agranoff and McGuire2003; Jacobson, Palus, and Bowling forthcoming; Moynihan forthcoming; Huxham Reference Huxham2000; Koppenjan and Klijn Reference Koppenjan and Klijn2004; Lundin Reference Lundin2007; Jokisaari and Vuori forthcoming; and Krueathep, Riccucci, and Suwanmala forthcoming).

We now also see the proliferation of theoretically driven efforts to explain network formation and structure—networks as dependent variable—as well as theories used to explain the outputs and outcomes of networked public action. Among the theoretical strands that offer some promise in this latter sense, for instance, there are inductively based multivariate explanations for collaborative success (for instance, some studies included in O'Leary and Bingham Reference O'Leary and Bingham2009); game-theoretic examinations of networked coproduction (Weissing and Ostrom Reference Weissing, Ostrom and Scharpf1993); public-choice-grounded analyses of common-pool resource management and use (Tang Reference Tang1991); agent-based models applied to the study of behavior in social networks that help shape policy results (for instance, Lazer and Friedman Reference Lazer and Friedman2007); transaction-costs approaches and institutional collective action (Feiock and Scholz Reference Feiock and Scholz2010); theory of policy implementation, grounded particularly in inter-organizational theory (O'Toole and Montjoy Reference O'Toole and Montjoy1984); and theoretical notions based upon the “structural holes” argument from sociology (Burt Reference Burt1992). The use of tools such as social network analysis, furthermore, provides an array of conceptual instruments and network characteristics that can aid in empirical analysis. And Meier and I have been at work in efforts to model formally the role of public-managerial action, with attention to—among other things—network aspects of the institutional setting as well as the networking behavior of public managers themselves. More about that in a moment.

PUBLIC ADMINISTRATION AND NETWORKS: WHAT WE KNOW AND WHAT WE NEED TO KNOW

So much for a few general comments on various research agendas. What I turn to now is the relevance of network and networking studies for public administration—and also for political science. My argument is that the proper study of public administration as a field cannot be undertaken without taking into account the networked character of much public action and the networking behavior of public managers. Furthermore, I want to suggest, research in the field of political science cannot help but be improved by devoting systematic attention to networks and networking in and around public administration. Whether it has been recognized or not, in other words, networks represent a tie that binds these realms of scholarship.

There are many ways to develop these points. I shall delineate a set of research questions and topics the exploration of which is, in many cases, already underway, and for which a network perspective is bound to be essential. I begin with some matters directly associated with public administration and then shift to some topics that pertain to political science.

Let us start here: Does the structure of networked public action and does the networking behavior of public managers matter? In particular, do these features of the public administrative setting influence the outputs and outcomes of public programs? A significant part of the research program in which Ken Meier and I have been involved is aimed at determining answers to these questions. The formal model from which we have built much of our research program incorporates network-related themes in two distinct places: in the partially structural term of the model, the “S” term, and in the so-called M ₂ term, that aspect of managerial behavior aimed at interacting with the “networked” part of the interdependent environment (O'Toole and Meier Reference O'Toole and Meier1999). The former may include network features and the latter involves managerial networking.

I limit my comments to a few brief bottom lines: Structural networks shape performance. Managerial networking, which is a managerial choice rather than a resultant of external pressures (Meier and O'Toole Reference Meier and O'Toole2005), does as well. The two sets of features interact in complicated ways in terms of their performance effects. For instance, managerial networking matters more to performance when core organizations operate in networks in which they are more financially dependent on others (O'Toole and Meier Reference O'Toole and Meier2004b). Managerial networking, which appears to be ubiquitous (for studies set in law enforcement, public education, and English local governments, see—respectively—Nicholson-Crotty and O'Toole Reference Nicholson-Crotty and O'Toole2004; Meier and O'Toole, forthcoming; and Walker, O'Toole, and Meier Reference Walker, O'Toole and Meier2007), tends to boost outcomes (Meier and O'Toole Reference Meier and O'Toole2001, Reference Meier and O'Toole2003), and not simply in a linear, additive fashion but also in a nonlinear way with respect to selected resources—yet not for all stakeholders (O'Toole and Meier Reference O'Toole and Meier2004a); networking is thus a manifestation of the intermixture of politics and administration. Managerial networking generates diminishing returns as well, although when managerial quality is taken into account the result disappears—at least in part because talented managers are more selective in using their time wisely (Hicklin, O'Toole, and Meier Reference Hicklin, O'Toole and Meier2008). The performance-related effects of such networking are boosted in the presence of enhanced managerial capacity, at least in situations where managers' organizations confront jolts or negative shocks from the networked setting (Meier and O'Toole, forthcoming; Andrews et al. Reference Andrews, Boyne, Meier, O'Toole and Walker2009b).

What impels the frequent use of networks and the adoption of networking behavior by public administrators? Here the evidence thus far is less systematic. But two sorts of answers can be given: one is based upon the literature and theoretical structure of the field; the other is built from empirical results of a number of studies of policy implementation carried out in complex institutional settings.

First, the most prominent theoretical justification for why standard-issue bureaucratic structures should be the most appropriate forms for public administration is actually a contingent one—one premised upon an important assumption regarding the nature of public problems that governments are called upon to address—and this contingency is being challenged by the nature of policy agendas.

Herbert Simon sketched the most compelling argument for bureaucratic organization available in our literature—his logic was based on the considerable advantages accruing to limitedly rational decision makers when the structure of organized action is designed to match the structure of important and recurring policy challenges. Thus are the advantages of standard operating procedures, patterns of directed search, specialization, near-decomposability of organizational structuring, and so forth. But Simon added a crucial qualifier that gets to the nub of the issue. In his words, “Fortunately, the problem of choice [that is, decision making] is usually greatly simplified by the tendency of the empirical laws that describe the regularities of nature to arrange themselves in relatively isolated subsets.” If this were not so, “even an approach to rationality in real behavior would be inconceivable.” (Simon Reference Simon1997, 78–79)

Now, if one were to confront Simon's contingency with the evidence that governments are being asked to address and manage so-called “wicked problems” (Rittel and Webber Reference Rittel and Webber1973)—challenges that ramify across putatively discrete sectors of policy—the issue is brought into clear focus. To the extent that our problem structures are messy (not built so as to be able to be handled in “relatively isolated subsets”), so too will be our institutional arrangements (Chisholm Reference Chisholm1995). On questions ranging from energy independence to health care to sustainable development, problems cannot be effectively addressed without drawing together organizations and actors from apparently disparate sectors and often very different levels in the governance system. In other words, networks, in some form, are required. But we do not have much valid theory to guide the organization or the management of such constellations of parties.

One could also approach the matter from another direction. Empirical examination of many public administrative settings indicates that the implementation of public policy through networked arrays is impelled by multiple forces. Without going into detail, and aside from isomorphic pressures, some are clearly political: adding actors to an implementation cluster to build support, especially in pluralist settings (Pressman and Wildavsky Reference Pressman and Wildavsky1984), or—on occasion—offloading some public responsibilities to sets of nonprofits when the state is interested in distancing itself from direct involvement in a controversial task; this latter impetus seems to have contributed to the current financial crisis (Cassell and Hoffmann Reference Cassell and Hoffmann2009, 616–17; Corder Reference Corder2009, 628). Some forces are even ideological—as with policymakers' preferences to contract out significant portions of government's business (Savas Reference Savas1987). Others are technical (and thus related to Simon's argument): drawing needed expertise, resources, and organizational capacity into the implementation process (O'Toole, Hanf, and Hupe Reference O'Toole, Hanf, Hupe, Kickert, Klijn and Koppenjan1997). Needless to say, these myriad forces are not likely to slacken anytime soon.

Now, to another question: Moving from the organizational to the network level, how do network structure and efforts to manage the networked array influence outputs and outcomes? Here we know far too little, and the challenge to making progress is daunting. Fully depicting complex networks for carrying out policy is labor intensive, and such efforts are typically restricted to small numbers of cases for a given study (for instance, Provan and Milward Reference Provan and Milward1995). Therefore, explaining dependent variables like outputs must be done without controlling for a range of other influences. Another difficulty is that although social network analysis has provided us with a rich conceptual apparatus for characterizing networks—multiplexity, density, various forms of centrality, and so forth—validated theory using such concepts to explain results is in short supply. And “management” in such arrays is a complex vector—many managers exerting effort variously to manage within their home organization, externally from their home organization toward one or more other network actors, and (for some) steering or influencing the network as a whole (O'Toole Reference O'Toole, Brudney, O'Toole and Rainey2000).

Beyond these complications, there is also the requisite complexity of theory to contend with. We have attempted to depict the influence of management (and other forces) on network performance, but the mathematics of this enterprise is contingent in part on the structure of interdependence within the network. Suffice it to say for present purposes that we have had to stipulate new mathematical aggregation operators to contend with the complexity (Meier and O'Toole Reference Meier, O'Toole, Lynn and Ingraham2004).

All in all, therefore, the answers to this question are critical, if we are to understand the world of networks and public administration, but they will not come easily.

Now, we could also explore the implications of what we have been discussing for several of the “traditional” subfields of public administration. To take but one example, public human resources management becomes a rather more challenging set of tasks if the people handling our programs are drawn from different, even disparate, civil service systems, not to mention the nonprofit or business sectors. And then there are the slimmed down, sometimes performance-contract-grounded organizations spawned in various countries by the so-called New Public Management. And what happens to public service motivation, in a cluster of interdependent organizational actors owned, managed, and likely motivated in disparate fashions across levels and sectors?

REACHING BEYOND: NETWORKED PUBLIC ADMINISTRATION AND SOME ISSUES OF POLITICAL SCIENCE

But rather than dig more deeply within public administration, we can also reach beyond to consider how the networked world in which public managers ply their trade might speak even more broadly—to some of the traditional concerns of political science. The theme here is straightforward: effectively addressing many of the core issues of political science requires that researchers take explicitly into account the networked character of public administrative action. Let me cite several such issues:

Political control of the bureaucracy. The conventional political science literature on control has it that political overseers seek to direct administrative agents, and research tries to show shifts by agents or their outputs when principals change or initiate new policy (for instance, Wood Reference Wood1988; Wood and Waterman Reference Wood and Waterman1993; Scholz and Wood Reference Scholz and Wood1999). But if the agents are typically multiple, the principal-agent model is misspecified from the outset (Hall and O'Toole Reference Hall and O'Toole2000). Political scientists have considered the matter of multiple principals—but not multiple, and interconnected, agents (for instance, McCubbins, Noll, and Weingast Reference McCubbins, Noll and Weingast1987). To compound the problem, any true test of political control would need to incorporate the values of the (possibly multiple) agents into the picture, not merely those of the principals, to test for control—but typically the agents are left out of the (literal) equation (Meier and O'Toole Reference Meier and O'Toole2006). In short, most of the research literature and received wisdom in political science regarding political control needs to be rethought, re-specified, and re-estimated to take into account multiple, networked agents who actually have schedules of values.

The relative efficacy of international agreements, especially multilateral ones. Here we move to the realm of international relations. As Ken Hanf and I have argued, increasingly public administration in the United States and elsewhere operates in a densely interconnected international system in which local decisions and actions may trigger global repercussions—and vice versa—and the fate of communities in one region is bound to the choices of decision makers elsewhere (O'Toole and Hanf Reference O'Toole and Hanf2002, 158). And we were not writing about the global financial crisis. Rather, we sought to analyze what I regard as a development of immense significance to both public administration and also international relations: the accretion of many international agreements among nation-states, agreements designed to address pressing policy problems that have transnational impacts. I am referring to the growing institutionalization of initiatives that are designed to facilitate the management of policy challenges beyond the control of any single state: problem-specific transnational networks grounded in multilateral formal agreements and, typically, managed in part via some version of a secretariat. These agreements have proliferated over the years and now touch virtually every field of policy.

The United States, for instance, is party to thousands of international agreements. Most of these are bilateral, but there are also many multilateral ones. In the field of environmental policy and management alone, a couple of hundred such agreements are currently in force, with approximately 26 added since 2005 alone. For the participating countries, the adoption of an agreement often triggers new or altered domestic activity, including by public administrators, and often at multiple levels in a country's governance system. And on-the-ground experience by public administrators can also stimulate upward pressure for changes in the structure or operations of such agreements. Beyond that, there is a managerial function performed at least partially by the international secretariat and aimed at achieving successful outcomes. In other words, the public management task for such agreements is a multi-level as well as multilateral one, and the (typically tiny) secretariats perform important functions but of necessity cannot do any of the managerial heavy lifting within signatory countries. In short, there is effectively no “top” to many of these functionally specific multilateral regimes. Rather, the secretariat assists the transnational negotiation and compliance system but the major portion of involved institutions and human talent is “retained in and by signatory states and their subnational elements. These are self-woven into and contribute to the shaping of the transnational order” (O'Toole and Hanf Reference O'Toole and Hanf2002, 162). The result is more complicated than what we have long considered the traditional administrative pattern, but the interdependence may make effective action more likely rather than less—depending on the details.

For social scientists who sit at the intersection of public administration and political science, the really interesting thing is that as these agreements are implemented, there is a considerable variance to be explained. Some cases are clearly ineffective. Some achieve significant behavioral change but fall short of full success (as with the Kyoto Protocol addressing global climate change). And some are striking successes. An example is the Convention on Long-Range Transboundary Air Pollution, or LRTAP, which ties together 51 countries on three continents. With no “top” and no effective supra-national regulatory or enforcement apparatus, this agreement—actually, it has been extended over time via eight additional pollutant-specific protocols—has been impressively effective (Haas, Keohane, and Levy Reference Haas, Keohane and Levy1993; Underdal and Hanf Reference Underdal and Hanf2000).

This set of developments suggests no end to states or their administrative apparatuses, but it does indicate the emergence of politico-administrative systems, multi-level ones, that sometimes bind many state-centered actors into enduring patterns of effective action despite the absence of authority flowing from the top. Understanding such transnational implementation networks, and the determinants of their success, should be a key research agenda item well into the future.

The role of social capital in shaping public life. For some decades now, but especially since Robert Putnam developed forceful arguments regarding the importance of social capital in political and social life, including its contributions to the strength of functioning democracies has been a salient topic (for instance Putnam Reference Putnam1993). Interestingly, it has not been much explored by scholars of public administration (an exception is Brewer Reference Brewer2003). But consider the implications of a networked public administration and of public administrators engaged regularly in networking behavior. Such behavior has the potential to leverage additional organization for the accomplishment of public purpose. Indeed, in work co-authored with Rhys Andrews, George Boyne, Ken Meier, and Richard Walker and tapping data on the performance and management of English local authorities, we have found that the drop in public sector performance in recent years attributable to unexpectedly huge waves of immigration from the so-called accession countries of eastern Europe—an influx 17 times larger than had been forecast—is mitigated by the density of community organizations in a given jurisdiction and the managerial capacity of the local authority (Andrews et al. Reference Andrews, Boyne, Meier, O'Toole and Walker2009b). We also have some preliminary findings that the determinants of managerial networking in these authorities includes the explanatory variables expected from structural holes theory (Andrews et al. Reference Andrews, Boyne, Meier, O'Toole and Walker2009a).

These aspects of the story draw in a theme from broader political science—social capital and its importance—to tell what is largely a management-and-performance story. What might connect this story back outward, to the core of political science, is if such impacts in turn carry explicitly political consequences. So, for instance, does the enhanced performance that managers facilitate help to suppress the electoral growth in popularity of the far-right, anti-immigration British National Party? We are currently at work on this question, which presumably should be of interest to still other political scientists: scholars of electoral behavior.

The mixed and varied way that the law shapes networked action. Here I can make the point exceedingly briefly, since Don Kettl's Gaus Lecture just last year (Kettl Reference Kettl2009) called attention to some of the implications regarding law—another of the traditional concerns of political science—for the conduct of public action, once that action is spread across jurisdictions and sectors. When public organizations, for instance, contract out part of their business to private entities (and some of these in turn subcontract with still others) and when they hive off portions of their responsibilities into privatized arrays, complex and confusing mixes of administrative law, contract law, and effectively no legal guidance toward accountable action may be interwoven in shifting patterns. So putting networked public administration into clear focus should cause a rethinking of whether and how public law controls and holds accountable those who are expected to achieve public purpose.

Public policy and American politics. The systematic study of public policy processes has sometimes incorporated network-related themes and perspectives (for a review of some of that scholarship as of a few years ago, especially as it pertains to public policy, see Robinson [Reference Robinson2006]; influential conceptualizations include, most obviously, Heclo [Reference Heclo and King1978] on “issue networks” and Sabatier and Jenkins-Smith [Reference Sabatier and Jenkins-Smith1993] on the advocacy coalition framework). Political science, furthermore, has certainly attended to interest groups, agency structures, and bureaucratic politics. However, networks place public managers in what can be expected to be a more active policy role, and political science thus far has devoted little attention to the activities and managerial processes used in networked settings as the managers of public organizations influence policy itself. Research on this subject, as well, should pay substantial dividends.

Cross-national comparative politics. It is clear from many studies in public administration that networks and networking by managers operate in many countries, not merely in a few. The networked patterns involving public managers may show some consistent patterns in terms of structure, behavior, and political as well as performance effects from country to country. Indeed, some of the work that we have performed using Texas school districts as units of analysis has already been replicated in the UK, with our British colleagues mentioned earlier. Currently, other efforts with additional colleagues are underway using Danish and Dutch public education as empirical settings. But of course political science tells us to be alert to the ways that different political systems differently shape results, and what we know about pluralist versus corporatist patterns suggest some potentially important differences in networks and networking. The literature of comparative politics offers raw material for theoretically grounded hypotheses about similarities and differences in network patterns across political systems. It is time, therefore, for some cross-national analysis in the best tradition of comparative politics but with a focus on the networked patterns of administrative actors.

Networks, public administration, and normative political theory. An additional segment of political science has only barely been influenced by the presence of networks for public administration, but the issues are important and the theme may become prominent.

We are used to considering public administration and normative themes, including those tied to democratic theory, in one of two ways. First, the received notion of political accountability for bureaucratic institutions is built upon the idea of political leaders sitting in authoritative positions and exercising oversight on administrative actors. This image is Emmette Redford's idea of “overhead democracy” (Reference Redford1969). We all understand that this perspective is an oversimplification, indeed that governments design additional mechanisms to open bureaucratic action to the scrutiny and advice of others—for example, through myriad mechanisms of stakeholder participation in the administrative process. But the theme of administrative accountability as it is sketched in the literature—particularly the political science literature—is predominantly of the top-down variety.

Second, there is normative theory with respect to what my colleague Barry Bozeman calls public values.Footnote ¹ Here we encounter the familiar analyses of such themes as efficiency, responsiveness, effectiveness, neutrality, transparency, access, justice, and equity, among others. And it is here that specialists in public administration have developed a number of normative arguments—from the theme of “politically neutral competence,” to use Kaufman's language (Reference Kaufman1956) to refer to a line of thinking reaching back at least into the Progressive Era; to the “social equity” argument embedded in scattered parts of what had once been called the New Public Administration (Frederickson Reference Frederickson1980); to the “agency perspective” embraced by the authors of the Blacksburg manifesto (Wamsley et al. Reference Wamsley, Bacher, Goodsell, Kronenberg, Rohr, Stivers, White and Wolf1990); to the themes of efficiency, customer service, and performance touted by the advocates of the New Public Management (Barzelay Reference Barzelay2001) and the proponents of public choice. Most of these normative perspectives are grounded fundamentally in some version of democratic theory in political science—although not particularly so in the case of the Blacksburg perspective, and in what is fundamentally an economist's version of democratic theory, in the case of public choice (Ostrom Reference Ostrom1989).

Consider now the implications for each of these two strands of normative theory in a world where public administration inhabits a social context of complex networks. First, with regard to the theme of accountability: What does accountability mean in a multi-governmental, multi-level system? Or one with public, private, and nonprofit organizations linked in patterns of nearly intractable interdependence? And how does one sort out the inevitable issues of diffusion of responsibility (O'Toole Reference O'Toole, Hanf and Toonen1985) as well as collective action when it is virtually impossible to establish causality, and when even the best efforts to date at performance measurement largely (perhaps inevitably) focus on organizational outputs and outcomes rather than network ones—PART scores notwithstanding? Some analysts might think the solution lies in an embrace of 360-degree accountability for performance (Behn Reference Behn2001), but I would suggest that this idea entails as many problems as it seems to solve. A prelude to careful analysis of the accountability challenge is to develop a suitable vocabulary and also a mode of analysis to discern the full range of issues at stake here.

Second, with regard to a consideration of public values, it seems clear that administering in a networked world touches upon many of our most salient public values in important ways that need to be analyzed; and choices about network design and networking patterns by public managers and others should be made after a careful consideration of such tradeoffs and consequences. Meier and I have detected a worrisome pattern, for instance—that managerial networking creates benefits for the more advantaged portions of a public organization's clientele but not for the more marginalized stakeholders (O'Toole and Meier Reference O'Toole and Meier2004a). And some of the British literature on networks points to a lack of transparency and accessibility in many such multi-actor systems in that country (Marsh and Rhodes Reference Marsh and Rhodes1992). Further, to the extent that one buys the argument put forward by the Blacksburg group that the so-called “agency perspective” is the crucible within which the long-term public interest is forged (Wamsley et al. Reference Wamsley, Bacher, Goodsell, Kronenberg, Rohr, Stivers, White and Wolf1990), networks would seem to dilute internal organizational decision making and deliberation—networks might even produce a postmodern babble in place of rational choice.Footnote ² Indeed, the regular use of networks to conduct the work of public administration might itself reshape and reorder just which public values are considered most important.

Allow me, then, to invoke and then modify a notion offered by Herbert Simon, who suggested that public administration should be considered a “design science” and who claimed that “the highest level of integration that” we “achieve … consists in taking an existing set of institutions as one alternative and comparing it with other sets.” This form of action, argued Simon, is akin to the analyses of “utopian political theorists” (Reference Simon1997, 111). If my argument is valid, then we could surely use a design science of and for networks. Such an enterprise would bridge the worlds of normative political theory and public administration (Sørensen and Torfing Reference Sørensen and Torfing2007).