Can Islands of Effectiveness Thrive in Difficult Governance Settings? The Political Economy of Local-Level Collaborative Governance

Many low-income countries contend with a governance syndrome characterized by a difficult combination of seeming openness, weak institutions, and strong inter-elite contestation for power and resources. In such countries, neither broad-based policy nor public management reforms are likely to be feasible. But are broad-based approaches necessary? Theory and evidence suggest that in such settings progress could be driven by "islands of effectiveness" -- narrowly-focused initiatives that combine high-quality institutional arrangements at the micro-level, plus supportive, narrowly-targeted policy reforms. This paper explores whether and how local-level collaborative governance can provide a platform for these islands of effectiveness. Drawing on the analytical framework developed by the Nobel-prize winning social scientist Elinor Ostrom, the paper reviews the underpinnings of successful collaborative governance. It introduces a simple model for exploring the interactions between collaborative governance and political economy. The model highlights the conditions under which coordination is capable of countering threats from predators seeking to capture the returns from collaborative governance for themselves. The relative strength in the broader environment of two opposing networks emerges as key -- "threat networks" to which predators have access, and countervailing "trumping networks" on which protagonists of effective collaborative governance can draw. The paper illustrates the potential practical relevance of the approach with three heuristic examples: the governance of schools, fisheries, and road construction and maintenance. It concludes by laying out an agenda for further empirical research, and suggesting what might be the implications of the approach for future operational practice.


Policy Research Working Paper 5842
Many low-income countries contend with a governance syndrome characterized by a difficult combination of seeming openness, weak institutions, and strong inter-elite contestation for power and resources. In such countries, neither broad-based policy nor public management reforms are likely to be feasible. But are broad-based approaches necessary? Theory and evidence suggest that in such settings progress could be driven by "islands of effectiveness"-narrowly-focused initiatives that combine high-quality institutional arrangements at the micro-level, plus supportive, narrowly-targeted policy reforms.
This paper explores whether and how local-level collaborative governance can provide a platform for these islands of effectiveness. Drawing on the analytical framework developed by the Nobel-prize winning social scientist Elinor Ostrom, the paper reviews the underpinnings of successful collaborative governance. It This paper is a product of the Public Sector Governance Unit, Poverty Reduction and Economic Management Network. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank. org. The author may be contacted at blevy@worldbank.org. introduces a simple model for exploring the interactions between collaborative governance and political economy. The model highlights the conditions under which coordination is capable of countering threats from predators seeking to capture the returns from collaborative governance for themselves. The relative strength in the broader environment of two opposing networks emerges as key-"threat networks" to which predators have access, and countervailing "trumping networks" on which protagonists of effective collaborative governance can draw.
The paper illustrates the potential practical relevance of the approach with three heuristic examples: the governance of schools, fisheries, and road construction and maintenance. It concludes by laying out an agenda for further empirical research, and suggesting what might be the implications of the approach for future operational practice.

Can Islands of Effectiveness Thrive in Difficult Governance Settings? The Political Economy of Local-level Collaborative Governance I: Introduction
The World Bank"s 1997 World Development Report, The State in a Changing World, signaled a new commitment among mainstream development scholars and practitioners to take governance seriously. At first, recognition of the critical role of governance seemed to point naturally towards a fairly straightforward (if ambitious) agenda of fostering "good governance" institutions. The general idea was that the institutional characteristics of capable and accountable states are well known --so the reform challenge was to redesign a country"s governance system to incorporate these characteristics.
But over the past decade, appreciation has grown as to the relevance of politics in shaping and constraining the governance agenda. 2 And when politics is brought into the equation, the implications of the insight that institutions matter become considerably more complex. Politics (stakeholders, and their power, incentives, skill, capacity to organize and constraints) inevitably shapes the dynamics of reform. A country"s economic, social and political institutions cannot be re-engineered from scratch. Nudging forward this interdependent, dynamic process calls for different approaches to engagementand different ways of identifying which approaches make sense across different country contexts.
The mainstreaming of governance and political economy into development thinking thus has profound and unsettling implications for both research and practice. It poses multiple challenges: a substantive challenge of revising our understanding of the drivers of development across different country settings in a way that takes both institutions and politics seriously; a policy challenge of finding new ways forward in the face of this revised understandingespecially in some of the more seemingly intractable governance settings; and a research challenge of building a program that is both analytically robust, and capable of providing a new generation of useful insights and guidance for development practitioners.
The principal purpose of this paper is to lay out a different approach to thinking about, and engaging in, these difficult governance environments. The suggested approach integrates the analysis of institutions, and the analysis of political economy, with a focus on the institutional arrangements for collaborative governance. Section II sets the stage by detailing some emerging findings from recent efforts to integrate 3 political economy into mainstream development thinking. Section III lays out the proposed approach in some detail. Section IV offers some concrete illustrations; explores the implications of the approach for a new generation of empirical research; and highlights its potential relevance for operational practice.

II: Institutions, Elite Bargains, and Development -Some Recent Research
When development practitioners take institutions seriously, one set of challenges emerges with special saliencenamely that many countries (generally, but not only, low-income; in Africa, and elsewhere) must contend with a governance syndrome characterized by a difficult combination of seeming openness, weak institutions, and strong inter-elite contestation for power and resources. 3 Characteristic features of this governance "syndrome" are:  Discretionary decision-making, and personalized deals among political elites;  Engagement of non-elites in the political process through clientelistic relationships, not through programmatic commitments to deliver better development performance;  The centrality of rent-generation, capture and distribution to both inter-elite bargaining, and political stability more broadly.
A recent series of path-breaking theoretical and empirical analyses on the interactions between institutions, political economy and economic performance shed light on the syndrome. At the aggregate level, key recent contributions are North, Wallis and Weingast (2009), North, Wallis, Weingast and Webb (2011, forthcoming), Bates (2008), and Khan (2010). A common theme in this literature is that in lowincome countries with underdeveloped formal governance institutions the locus of political stabilityand hence credible commitment, and private sector developmentis to be found in negotiated bargains among political elites. These elite bargains center around mutually beneficial agreements vis-à-vis the generation and distribution of rents. Bates rigorously models the decision as to whether to cooperate or turn to violence in a game-theoretic format. As Khan details, in some settings elite bargains are centrally coordinated by dominant leaders and/or tightly-managed political parties; in other settings, the bargain involves more flexible coalitions, with greater attendant risks of moral hazard.
Viewed from the perspective of the normative literature on country-level governance, the implications of this work are bleak: Over the short-to medium-term the prospects are vanishingly small of transforming negotiated elite bargains into beacons of good governance. Neither the public bureaucracy nor the 4 institutions that are expected to hold it to account are likely to work well. Policy will be politicized; topdown, system-wide approaches to implementation are likely to be ineffective. Given the prevailing incentives, neither broad-based policy nor public management reforms are likely to be feasible.
But are broad-based approaches necessary? The development experience of a country such as Bangladesh (which, as documented in Khan (2011) has combined strong poverty reduction and steady growth with endemic corruption, clientelism, and polarized, seemingly dysfunctional political contestation) demonstrates that development can move forward even in the absence of broadly encompassing "best practice" reforms. Recent work, summarized in Rodrik (2007), highlights how relatively modest reforms which alleviate "binding constraints" can unlock growth. What appears to drive progress in these settings are "islands of effectiveness" --narrowly-focused initiatives that combine highquality institutional arrangements at the micro-level, plus (as necessary) narrowly-targeted policy reforms that facilitate the emergence and operation of these "islands". What might be the political economy underpinnings that make it possible for such micro-level institutional initiatives to take root?
A variety of contributions that address the political economy interface between the aggregate and microlevel offer some insight:  In an early effort, Levy and Spiller (1997) explored the relationship between regulatory arrangements, national checks and balances institutions, and performance in the telecommunications sector across a half-dozen countrieshowever, viewed through the lens of the more recent contributions, their analysis of the micro-level drivers and constraints focuses excessively on formal institutions.
 Keefer and Khemani (2005) examine the ways in which clientelistic political systems have a bias towards the provision of highly-targeted rather than broad-based public goods. Two contributions by Stephen Haber and associates are especially illuminating. Haber et al. (2003) detail the institutional arrangements that underpinned sector-level dynamism in Mexico"s clientelistic political economy almost throughout the half-century from the 1880s-1930s. Haber et al. (2008) extend the analysis for middle-income Mexico over the subsequent half-century. A central finding in these analyses is that, contra the general view of property rights as a public good, they can in fact be "privatized". Haber explores both theoretically and empirically the role of ad hoc discretionary institutional arrangements -partnership, monitoring, and (in a game-theoretic sense) "hostage-taking" --that link individual private investors with politically influential actors who are well-positioned to assure that government does not renege on its promises. Such arrangements, he shows, can provide the credible commitments private actors need to invest, even during difficult, violent times.
Haber"s conclusions are profoundly discomfiting for policymakers, donors and other development practitioners who focus their attention on "best practice" solutions. But viewed from a different perspective than "best practice", they can also be interpreted as pointing towards a way forward in difficult settings: The results suggest that, beyond a very de minimis agenda, growth in weak governance settings does not require good broad-based policies and institutions. More focused policy reforms, and more narrowly targeted institutional arrangements can suffice. This is a key motivating insight for the approach laid out in this paper.

III: (How) Can Collaborative Governance Be Effective in Difficult Governance Settings? A Framework
This section explores analytically the potential role of local-level collaborative governance as an entry point for development in weaker governance settings. First, it highlights the relevance of collaborative governance-both as an approach that potentially can be applied in a wide variety of settings, and as a mechanism for limiting the costs of dysfunctional governance. Thereafter it reviews analytically the underpinnings of successful collaborative governance, and introduces a simple model for exploring the interactions between collaborative governance and political economy in countries where formal institutions are weak.

III.A: Varieties of Collaboration and Their Hazards: Concepts and Examples
Collaborative governance comprises a set of institutional arrangements designed to achieve a collective purpose in which there are multiple principalsincluding principals from the public sector. As Box 1 clarifies, there is a very close correspondence between the definitions of collaborative governance and of collective action. Collective action, though, has come to be associated with situations where the relevant principals are non-governmental, and also sometimes is applied narrowly to the governance of common pool resources. Though this paper draws liberally on (and extends) the literature on collective action, the term "collaborative governance" is preferred as a way of making explicit that the framework is intended to have broader applicability than is usually associated with collective action.
Box 1: Collaborative governance and collective actionsome definitions Ostrom (1990) defines "collective action" as the process whereby "a group of principals who are in an interdependent situation can organize and govern themselves to adopt coordinated strategies to obtain (and maintain) higher joint benefits when all face temptations to free-ride, shirk, or otherwise act opportunistically…" The process is one in which a new set of institutions [rules of the game] that facilitate cooperation are supplied, in which participants credibly commit to follow the rules, and in which principals engage in mutual monitoring of conformance to the rules…..
A striking feature of the above definitionand the reason that the term "collaborative governance" is preferred in this paper --is that, in principle, the relevant principals can comprise both governmental and non-governmental actors. Including governmental actors as a co-equal among the principals presumes that, in the context of the specific collaborative endeavor being considered, they have similar standingas co-principalsas do nongovernmental actors. This is consistent with a legislated governance framework that delegates decision making to the collectivity. It also is consistent with the North, Wallis and Weingast (2009) approach to the analysis of governance in low-income and institutionally weak settings as a series of elite bargains; they argue that the classic characteristic of the Weberian state -"a monopoly over the the legitimate use of violence"is achieved relatively late in the development process. to initiatives among groups of private firms to provide quasi public goods; to collaborative governance of public service provision; and governance of autonomous public agencies.

Box 2: Varieties of collaborative governancesome examples
This paper takes a broad view of collaborative governance as a mechanism for creating and/or managing rents to achieve social rather than private purposes: (i) It incorporates (as per the discussion of Box 1) collaborative multistakeholder governance involving both governmental and non-governmental actors as well as action exclusively among non-governmental actors. (ii) It defines broadly what might qualify as "rents" that potentially are governable through collaborative mechanisms. The crucial commonality is that the realization of social returns is based on cooperation among diverse stakeholders who, in the specific context, operate largely as peers ("principals") vis-à-vis one another. Three distinct types of collaborative governance can be identified:  Collaborative governance that aims to govern common pool resources in a sustainable manner. This is the classic collective action challenge explored in depth by Elinor Ostrom. Examples from work by her and her associates include o Collective governance of irrigation systems; o Collective governance of inshore fisheries; o Collective governance of communally-owned land; o Collective governance of groundwater resources; o Collective governance of forests. As Ostrom details, these common pool resources can be managed through self-organization among nongovernment principals with access to the resource, through government regulation, or through hybrid arrangements involving partnership between governmental and non-governmental actors.
 Collaborative governance that aims to facilitate coordination among non-governmental actors to both produce a quasi-rent (a quasi-public-good) and share the benefits among all participant/beneficiaries. Examples include: o Cluster initiatives among private (manufacturing and other) firms within a given sector or value chain to facilitate mutually-linked investments, or to jointly invest in learning or market facilitation; o Outgrowing arrangements in which processers and small-holder farmers collaborate around common rules governing the provision of inputs and the sale of crops; these can combine vertical (i.e. between farmers and processers) and horizontal (i.e. between groups of farmers and groups of processers) institutional arrangements. Note that these arrangements might include some involvement by public actors in support of the private initiatives, and might include also advocacy of targeted policy reforms.
 Collaborative governance that aims to assure that public resources allocated for a specific service provision purpose are indeed used as intended. Viewed from the perspective of the potential beneficiaries, these public resourcese.g. government paid positions, or public financial resources made available to procure necessary inputs --can be interpreted as a rent/common pool resource with the potential to be used for the social purpose, or captured for private ends. Examples of collaborative governance in this domain include: o Collaborative governance of schools, health clinics and other front-line service provision facilities by multi-stakeholder groups which include service recipients and others with a stake in the efficiency and effectiveness of service provision. o Multi-stakeholder oversight of public sector procurementfor example through joint commitments by private bidders and public agencies to govern specific large-scale, procurementintensive public projects through "integrity pacts". o Public-private partnerships and/or other collaborative, multi-stakeholder arrangements to govern the operation of formally state-owned entities, and of other arms-length public agencies.
Though the examples in Box 2 are diverse, they share a common set of coordination challenges:  The ubiquitous challenge of facilitating cooperation among participants to achieve joint benefits, in a way that limits the classic free rider and other moral hazard challenges associated with collective action. [The term CA-Q is used below to designate this first type of productivityenhancing collaborative initiative.]  A challenge that is relevant especially in weaker governance settingsnamely coordination to counter threats from predators seeking to capture the returns from collaborative governance for themselves. [The term CA-P is used to designate this second type of predator-repelling Predation is a key concept for the analysis that follows. Box 3 below defines 'predation' in terms of the willingness and ability of predators to override with impunity formal and informal rules. Predators are individuals or groups who seek to capture collectively created or managed rents for private purposes. The threat from predators can manifest at either or both of two levels:  At the level of the collaborative initiative itself, predators might choose to ignore with impunity mechanisms agreed among participants for monitoring and enforcement, as not applying to them.
Additionally,  Predators might leverage their influence networks to put themselves out of reach of any formal legal frameworks to which CA-Q participants might otherwise have had recourse for resolving disputes and sanctioning illegal acts.

Box 3: Predation defined
Predation is defined here to refer specifically to actions that use channels of political support external to the specific arena of cooperation to override with impunity the formal and informal rules of the game associated with the collaborative effort. Note that predation is more far-reaching in its impunity than each of the following, common hazards of collective action:  Distributive conflicts among participants in a collective action effort as to how to share the benefits, and/or conflicts over who will lead the effortwhich need involve no element of opportunism. 4  'Free riding' in which a presumptive participant in a collective effort chooses to shirk on his obligations, but nonetheless enjoys a share of the benefitsmutual monitoring is a key way in which protagonists of collective action mitigate this risk.  Corruption, in which a participant in a collective action effort pays (or accepts) a bribe to illegally override an agreed-upon formal or informal rulebut (contra to predation) is vulnerable to detection and, if detected, is subject to sanction. .

III.B: How Collaborative Governance and Predation Interact
A central hypothesis of this paper is that productivity-enhancing and predator-repelling collaborative initiatives can be provided as joint products. Local level actors cooperate to produce a joint benefit and, in the process, are able to co-produce predation deterrence at low marginal cost. This is the (hypothesized) reason why otherwise unexpected islands of success can thrive in weak governance settings.
How do collaborative governance and predation (and its deterrence) interact with one another? The point of departure for the analysis is Elinor Ostrom"s Institutional Analysis and Development (IAD) framework (see Ostrom, 2005 for a comprehensive presentation). The framework links systematically the social goal being pursued, individual decisions as to whether to cooperate, the quality of collective action, and the results achieved. Consistent with rational choice theory, it anchors its analysis of cooperation in individual decisions, based on individual assessments of the costs and benefits --and likelihood of achieving the benefits --of cooperation. Figure 1 below provides a summary illustration of the connections. As the figure suggests, the individual decision as to whether or not to cooperate will be 9 influenced, in part, by the actual and perceived behavior of others with whom the cooperative endeavor might be undertaken.

Figure 1: How individual incentives, collective action and performance interact
Ostrom identifies a set of eight "good practice" principles for the governance of collective action. (See Box 4.) These good practice principles are derived from both theory and empirical analysis. Their theoretical anchors comprises the IAD"s "action situation" framework for analyzing institutionswhere action situations comprise "the social space where participants with diverse preferences interact, exchange goods and services, solve problems, dominate one another, or fight"; as per the framework, all action situations are governed by seven sets of "working rules" 5 Their empirical basis comprises the voluminous analysis by Ostrom and associates, over the past thirty-plus years, of the "working rules" which govern hundreds of common pool resources, in dozens of countries. On the basis of a metaanalysis of this work, Ostrom concludes that "robust systems for governing common pool resources had met most of the good practice principles, and that those systems that had collapsed or were performing ineffectively were not so structured" (Ostrom, 2005, p. 259).
An important gap in the IAD framework is that it does not explore the interactions between collaborative governance and the broader political economy environment. Instead, the framework treats politics as an 5 Ostrom (2005), p.14. Viewed through the lens of the IAD framework, the term 'institution' is thus defined precisely -more so than the usual, and very general, 'rules of the game' -as 'a set of working rules that govern an action situation'. The seven elements Ostrom identifies seven common elements of all action situations, each with an associated set of working rules. The seven elements are (i) actors and associated boundary rules which define who is eligible to enter a position; (ii) positions and associated position rules which create positions for participants to enter; (iii) actions and associated choice rules: which specify what a participant occupying a position must/must not/may do at a particular point in a decision process; (iv) control and associated aggregation rules which determine how collective decisions are to be arrived at; (v) potential outcomes and associated scope rules which define and regulate the range of acceptable outcomes (vi) information about actions and outcomes and associated information rules which assign the obligation/permission or prohibition to communicate to participants in positions… and the language/form in which the communication will take place; and (vii) net costs and benefits and associated payoff rules which assign rewards or sanctions. exogenous variable whose impact is derived via the mediating impact of higher level institutions. Thus, Ostrom (2005;p.58) suggests that "what can be done at one level will depend on the capabilities and limits of the rules at that level and at a deeper level". The dilemma, though, is that in low-income countries with weak institutions --the focus of the present paper --these higher-level arrangements are ad hoc, personalized, and discretionary. As generally applied, the IAD framework thus assumes away the political economy complexities that are central to the development challenge of many low-income countries.

Box 4: 'Good Practice' Principles for Collaborative Governance
This formulation of the principles is taken from Ostrom (2005Ostrom ( , 2009, adapted to fit the broader focus of the present effort on multiple types of collaborative governance, beyond the specific issues associated with common pool resources. In difficult-governance settings, what happens when productivity enhancing collective action (CA-Q) at local levels collides with predators seeking to capture these rents for themselves? Certainly, the presence of predators who aim to over-ride with impunity efforts to build rule-boundedness complicates the challenge of fostering collaborative governance. Butand this is the central proposition of the present papertheir presence need not render collaborative governance impotent. Protagonists of CA-Q potentially can resist the threat of predation by drawing on their own channels of influence. Thus:  "Threat" resources comprise the influence networks external to the action situation on which predators might draw to over-ride with impunity rules intended to facilitate cooperation and restrain discretion.
 "Trumping" resources comprise the countervailing influence networks external to the action situation on which participants in CA-Q might draw to facilitate compliance with rules.
Incorporating "threat" and "trumping" variables into the analysis provides an analytic bridge between Ostrom"s framework and the broader political economy considerations laid out in Section II. 6 Analytically, predation can be viewed as a "sub-game" within a broader extended cooperative game.
Potential predators have to decide whether to predate; participants in the prior CA-Q game have to decide whether to resist. To illustrate the logic of the situation, consider a scenario where: (i) quasi-rent enhancing cooperation (i.e. CA-Q) already is in place, and is functioning effectively; (ii) predators arrive on the scene, seeking to capture the rent resource; and (iii) the participants in CA-Q choose to resist. As the decision tree in Figure 2 suggests, the result will depend on the relative magnitude of threat and trumping resources.
Underlying the construction of Figure 2 are assumptions as to what information is available to the protagonists of the sub-game. If both CA-Q participants and predators are fully informed as to the magnitude of the threat and trumping resources that each has available, then (contra to Figure 2) conflict will never be observed. If "trump" > "threat", potential predators, realizing that they will lose, will refrain from engaging, and CA-Q will be effective and undisturbed. Conversely, if "trump" < "threat", CA-Q participants will conclude that it is pointless to resist, CA-Q will disintegrate, and performance will decline rapidly.

Figure 2: Predation and its Consequences -A Decision Tree
6 This analysis of conflict through the lens of a game structured around the relative power of threat and trumping networks parallels a discussion of the allocation of property rights in Khan (2000)  In practice, insofar as both "threat" and "trumping" operate through informal channels, knowledge as to their actual value is likely to be uncertainand also quite dependent on the specific identities and networks of individual playerssuggesting that conflict scenarios could be quite likely. Table 1a summarizes the options in a scenario where "trump" > "threat", but this reality is not recognized by one or both of the protagonists; Table 1b summarizes the converse scenario --the reality where "trump" < "threat". In all, four possible patterns are evident:  The "full information equivalent" scenario where perceptions are aligned with underlying reality, and the result is as described in Figure 2.
 Conflict scenarioswhere expectations are misaligned on the part of one of the protagonists, with conflict the immediate consequence, and a better-informed equilibrium outcome the eventual result.
 Unnecessarily poor performance scenarios (in Table 1a) --where predation dominates, but CA-Q protagonists could prevail, were they only willing to organize and take up the fight.
 Fragile, high performance scenarios (in Table 1b) --where CA-Q is effective, but could readily be reversed were predators to recognize their coercive potential.

Threat > Trumping
Limited/no observed collaborative governance A final figure illustrates the links among individual decision-making, the predator-threat sub-game, and the quality more broadly of collaborative governance across a variety of seemingly separate arenas.
Recall that (as per Figure 1) individuals base their decision as to whether or not to cooperate in part of their perception as to the probability of the benefits of cooperation being achieved. Figure 3 underscores that this decision is centrally influenced by their judgments as to the likely outcome of any predation threat. If they judged that the threat could be deterred effectively, individuals might choose to cooperate.
But if their judgment was that predation was too powerful to overcome, potential participants would recognize that the probability of enjoying the benefits of collective action was low, and so would be deterred from embarking on this always (even in the most benign circumstances) challenging endeavor.

Figure 3: The interaction between predation, individual expectations, and cooperation
While Figure 3 focuses on individual decisions, it points to the possibility of feedback effects from the broader environment, and hence of "tipping points" across a broad variety of arenas for potential cooperation. As the figure underscores, individual actors decisions as to whether to participate in a cooperative endeavor are based on their expectationsand these expectations are formed from many influences, indirect as well as direct. Pessimistically, in the face of repeated evidence of successful predation across their society, potential cooperators might increasingly give up on their efforts, resulting in a growing number of failures, and a downward spiral of quiescence and predation. But the alternative also is possiblea virtuous spiral of accelerating cooperation and enhanced performance across multiple arenas as an increasing number of hitherto quiescent local actors, inspired by an increasing number of brave successes, step forward to work together in pursuit of social goals.

IV: From Theory to Empirics and Practice
As noted at the outset of this paper, the motivation for exploring the analytics of predation is a cuttingedge real-world challenge: finding feasible ways of achieving development gains in settings characterized by a difficult combination of seeming political and economic openness, weak institutions, and strong inter-elite contestation for power and resources. The theory laid out in Section III provides a point of departure for inquiry. But generating knowledge that can help address this challenge calls for moving beyond theory and engaging also with the empirical reality.

IV.A: Three Heuristic Illustrations
This sub-section offers extended illustrations of the potential interactions between collaborative governance and the threat-trumping game for three of the many examples suggested in Box 2.

Multi-stakeholder school-governing bodies and teacher absenteeism.
Teaching positions allocated to public schools can be interpreted as akin to a "common pool resource" available to the local communitybut in many developing countries teacher absenteeism (free riding on the resource) is a ubiquitous problem. "Transparency and participation" increasingly are advocated as a solution --but what is the "transmission mechanism"? The analytic framework laid out in Section III suggests that there are two key dimensions. The first dimension comprises the quality of participation of parents and other stakeholders in school governance: Better quality engagement is hypothesized to provide a first line of defense against teacher absenteeismwith the "good practice" principles offering a benchmark against which to assess the quality of participatory school-level governance.
The second dimension comprises the threat-trumping dynamics. In difficult governance settings, teacher absenteeism is likely to be underpinned by powerful influence networks (e.g. teachers unions, or patronage-dispensing politicians). Well-organized school governing bodies can try to countervail the threat by drawing on their "trumping" influence networks (e.g. appeals to respected community and religious leaders, or to national NGOs). The analytic framework suggests that school performance in generaland the acceptance of absenteeism versus disciplinary action in particularis likely to be dependent on the relative strength of these contrasting networks.
Fishing rights. The allocation of valuable fishing rights often raises difficult distributive challenges; collective organization of a fishery has been one way of jointly addressing the challenges of equity and of sustainability. But, as per the definition in Box 1, contrast these distributive challenges with predationsay a foreign company that bribes officials to grant it rights that result in over-fishing in excess of a preagreed ceiling, and leverages its external influence networks (its "threat resources") to assure that no effort is made to overturn the regulatory decision.
How might incumbent fishing operations respond to this threat of predation? The analytic framework directs attention to the role of "trumping networks". Fishing companies seeking to protect both their continuing access to the resource, and its sustainability, might usefully seek to build alliances with environmental activistsor with retailers committed to sustainable trade --to expose and reverse policies that threaten overfishing. Globally, the Marine Stewardship Council (MSC), which offers "eco-labels" to fisheries that are certified as being managed sustainably, offers a natural potential ally. But why do only some fisheries, but not others, successfully leverage the MSC"s potential? The analytic framework suggests that the answer may lie in a combination of the readiness of fishing companies to work collaboratively to address these challenges -and in the relative strength of the "threat" and "trumping" networks.
Road fund governance. Across Africa, and elsewhere, there has been widespread creation of road funds as a way of ensuring that vehicle taxes, and other earmarked revenues, are used as intended for road construction and maintenance. But procurement contracts associated with road construction are notoriously corrupt. In institutionally weak settings, can collaborative governance of road funds protect against the predation of this public resource?
The analytical framework in Section III suggests that inclusion in the governing board of the road fund of users of the road network who have a stake in its quality (e.g. haulage companies, business associations) has the potential to counterbalance the corrupt use of road fund resources. But realizing this potential will depend (again) on two dimensions. First, are the collaborative governance arrangements themselves robustin the sense that the stakeholder participants genuinely are empowered as full participants in the corporate governance arrangements (as opposed to, whatever the formal rules might suggest, serving only at the pleasure of key political and bureaucratic actors)? Second, insofar as they are empowered to participate, what the strength of the "trumping" networks to which stakeholders have access relative to the strength of the predatory threat? The framework suggests that whether road funds have the potential to be effective in protecting against predation will depend on the answers to the above two questionsand that the answers almost surely will vary from country to country.

IV.B: An Approach to Empirical Research
This section will consider three questions as to how to apply the framework in empirical research: What sub-sample might most usefully be the target of empirical research? What research methodology is best fitted to the research task? What are the specific refutable hypotheses around which the empirical effort might be organized? The discussion will frame its suggestions as to how empirical research might proceed in quite general terms; the reader might usefully provide more concreteness by filtering the discussion through the lens of one or more of the specific examples laid out above.
Sub-sample selection. The theory laid out in Section III suggests that the interaction between collaborative governance and predation could result in three distinct empirical patterns:  Pattern I: No observed cooperation (and hence the absence of the good practice principles in the relevant action situation), poor development outcomes, and evidence of capture by one or more groups of predators.
 Pattern II: High quality collaborative governance (as evidenced by the presence of Ostrom"s good practice principles), good development outcomesand a seeming absence of predation (which, it was hypothesized, might signal either the absence of predators, or a perception by putative predators that trumping resources were more powerful than their threat resources, and thus that the returns from the effort to predate would be negative).
 Pattern III: Significant episodes of conflict between the protagonists of collaborative governance and predatorswith a wide variety of potential outcomes.
Pattern I is empirically uninteresting: with no cooperation evident, only governance failure can be observed. Pattern II is somewhat more interesting insofar as it can provide an empirical basis for exploring empirically Ostrom"s suggested relationship between the presence (or absence) of the "good practice" principles for collaborative governance, and development outcomes in areas other than the common pool resource problem on which the bulk of research has focused. Pattern III is most interesting: empirical research into these episodes of conflict holds high potential for contributing at the frontier of development research and practice. Indeed, for reasons laid out further below, it is proposed that the principal research effort be focused on Pattern III-style examples.
Research methodology. The "analytic narratives" framework, summarized in Box 6, is proposed as one which fits best with the empirical questions raised by Pattern III. Analytic narratives provide a disciplined approach for learning about the interaction between efforts at collaborative governance and associated threat-trumping sub-games in specific, concrete settings: They can shed light on both the proximate determinants of performance of the specific action situation being studied and, more broadly, on the character of influence networks that potentially extend all the way from the front-line to the national level. Exploring these interactions against the backdrop of knowledge as to how the conflict was resolvedi.e. who prevailedwill provide invaluable insight into the relative strengths of the competing networks in specific settings.

Box 6: The 'Analytic Narratives' Research Methodology -Some Key Features
Rational choice, with game theory as an analytic platform: "In an effort to move from [description] to explanation, we move from "thick" accounts to "thin" forms of reasoning. We seek to highlight and focus upon the logic of the processes that generate the phenomena we study. In doing so, we use rational choice theory. We find game theoretic models particularly useful ways of exploring the validity of narrative accounts….We seek to construct the game that provides the link between the prominent features of the narrative and its outcome". (p.14) Linking games and empirical research. "We seek to account for outcomes by identifying and exploring the mechanisms that generate them….By reading documents, laboring through archives, interviewing, and surveying the secondary literature, we seek to understand the actors" preferences, their perceptions, their evaluation of alternatives, the information they possess, the expectations they form, the strategies they adopt, and the constraints that limit their actions. We seek to cut deeply into the specifics of a time and place, and to locate and trace the processes that generate the outcome of interest." (pp. 11-12) Deduction and induction. Analytic narratives "blur the conventional distinction between deduction and induction…. The(ir) construction is an iterative process…They depart from conventional notions of hypothesis testing. The dominant response to disconfirmation is reformulation not falsification…. We move back and forth between interpretation and case materials, modifying the explanation in light of the data, which itself is viewed in new ways, given our evolving understanding …[so that] in the end we achieve a match between theory and case materials". (pp. 16-17) 7 ":  Generating theory. "Initially the theory is formed from the data; it is selected because it appears to offer a good fit. Rendered explicit, the theory then becomes vulnerable; it can be subject both to logical appraisal and to empirical testing.  Disciplining the narrative. "Theory places constraints upon the narrative; the account is constrained by the logic of the theory.  Modifying theory . "The cases derive implications from theory; but when the case materials do not confirm their expectations, the authors respond by reformulating their models and by altering the way in which they think about the problem.  Postdiction? "When models highlight features of the data that hitherto have escaped attention; when they can be contradicted by the evidence; and when they predict relationships that must hold, if their equilibria capture the processes that generate the phenomena of concernthen we are well beyond mere exercises in "curve fitting". Testable hypotheses. While it is in the nature of the "analytic narratives" approach that much of the learning about the character and relative strength of various influence networks will be inductive, five sets of initial hypotheses are laid out below as a way of clarifying the analytical target of the empirical endeavor, and thereby providing an initial guide to the (iterative) empirical work.
 H1. The presence of collaborative governance helps combat predation: 7 The authors underscore that their approach is consistent with the insights into the scientific method provided by Thomas Kuhn (1962) and Imre Lakatos (1970).
o H1A: Collaborative governance arrangements to sustainably manage rents, or to create quasi-rents (i.e. CA-Q), are likely to be successful insofar as their working rules are well aligned with good practice principles. o H1B: Effective CA-Q provides a platform for combating predation, insofar as it enables protagonists to combine their individual, potential "trumping" networks into a common predation-fighting front.
Research to address this hypothesis will provide the empirical baseline for the overall analytical narrative.
Key elements would include: (i) a description of the "outcome", that is the extent to which the collaborative governance initiative being studied has achieved its intended social goal; (ii) characterization of the working rules governing the collaborative governance situation being studiedwith the specific goal of assessing the extent to which they align with the "good practice" principles laid out earlier; (iii) documentation of what, if any, predation threats arose, and how they were resolved. This part of the "analytic narrative" would: (i) describe how the CA-Q cooperation emerged: how it was initiated, how it was consolidated, and how it was sustained, with a particular focus on the role of key individuals in its evolution; (ii) detail how the participants in CA-Q responded when faced with a threat of predation, with a specific focus on (iii) the role of key individuals in persuading the CA-Q participants to confront the threat, and in orchestrating the subsequent response.
 H3: Access to trumping networks helps combat predation: o H3A: There will be variations across CA-Q action situations in the extent to which the CA-Q protagonists have access to "trumping" networks of influence. o H3B: Settings where the CA-Q protagonists have better access to such networks -by virtue of higher incomes, histories of political activism, religious networks, or other context-specific linkagesare more likely to prevail over predators. (ii) assess the extent to which they were narrowly-focused policy initiatives or broad-based measures to facilitate the relevant enabling environment; and (iii) describe the origin of these policy measures. Of specific interest will be the timing of policy reforms and of the collaborative endeavor: which preceded which? Did policy reform catalyze cooperation, or did the protagonists of the relevant collaborative initiative successfully lobby for the requisite reforms?

IV.C: Implications for Practitioners
Operational practice rarely, if ever, waits on the results of robust research before moving forward. Indeed, a major push has been under way for some years to foster transparency, participation and accountability at the service delivery front-line (and efforts to foster inter-firm "clusters" and local-level collective action in the management of common pool resources are long-standing). What might be the implications of the approach laid out in this paper for enhancing these operational agendas?
First, and most obviously, the relevance of Elinor Ostrom"s IAD "good practice" principles potentially is much broader than the common pool resource arena where they usually are applied. As this paper has underscored, there is a rich variety of arenas of collaborative governance; for many of them, the good practice principles can provide guideposts for effective design. A broader application of the good practice principles may be especially helpful in giving more precise operational content to at least some aspects of 20 "participation"which has become ubiquitous, almost a panacea, in development discourse and practice but often is framed in exceedingly general terms.
Second, the analytical framework suggests that there may be some (perhaps many) arenas where protagonists of potential collaborative governance have not come forward in part because they perceive the threat of predation to be too high to be countered. However, given skillful leveraging of the latent trumping resources to which the CA-Q protagonists potentially have access, this perception could prove to be erroneous. Both cooperation and, by extension, the mobilization and orchestration of "trumping" resources are learned skills. So in principle practitioners might be in a position to provide practical guidance to CA-Q protagonists as to how they might more effectively engage. Such guidance calls for knowledge as to the details of how the "threat-and "trumping" mechanisms operateand these details are likely to vary widely from country-to-country (and, plausibly, across the regions and sectors within a country). There is thus substantial potential for synergies between practitioners and researchers.
Third, in those settings where the magnitude of the predation threat is larger than available trumping resources, it may be possible for practitioners to augment trumping resourcesand thereby reverse an earlier equilibrium. Two illustrative examples of potential augmentation:  Enhanced transparency: Transparent information on performance potentially provides CA-Q protagonists with a focal point around which to organize. Further --in the spirit of "transparency is the best disinfectant" --it has the potential to expose to the light of day, and thereby deter, practices which are broadly perceived as socially illegitimate.
 Support for strengthened horizontal associations. Insofar as protagonists of CA-Q are likely to vary in the extent to which they have access to trumping resources, stronger horizontal associationse.g. associations of school governing bodies or parents associations; farmers associationscould help close the gap, by providing new sets of allies to which CA-Q protagonists might turn for support.
One final point. Evidently, facilitating more effective collaborative governance involves quite substantial transactions costs, with the benefits likely to be "retail"one arena at a time. A skeptical reader might thus wonder whether the returns might not be higher from a more direct focus on broad-based policy reforms or broad-based efforts to improve public management. But recall the point of departure of this paper: Broad-based policy and public management reforms often are not likely to be feasible in difficult governance settingsbut focused collaborative governance initiatives (and complementary narrowlyfocused policy initiatives) do not require "good governance" as a pre-requisite, and both theory and 21 evidence has demonstrated that broadly encompassing "best practice" reforms are not necessary to move development forward. And recall also the possibility (illustrated in Figure 3) of "tipping points"of social learning in response to evidence of successful collaborative governance setting in motion virtuous spirals of accelerating cooperation and enhanced performance across multiple arenas by an increasing number of hitherto quiescent actors.
All too often, development advocacy in weaker governance settings comprises the endless reiteration of "best practice" policy and governance nostrums, which fail to be implementedwith the only fruits of the effort being frustration, hostility, cynicism and despair. In such settings, the effort to leverage collaborative governance to build islands of effectiveness emerges as a potentially more implementable alternative: It works "with the grain" of the prevailing political economy realities, and is capable of achieving results which, though often seemingly quite modest, are valuable in themselves. Further, the steady accumulation of small gains can help sustain hope as to the possibility of building a better life for all. Though the focus of individual efforts might be narrow, the overall impact need not be small.