Network Governance Analysis
- Introduced by
- Keith G. Provan & Patrick Kenis
- Year
- 2008
- Type
- Interorganizational network analysis framework
- Family
- Governance and collaboration network analysis
What does it do?
Network governance analysis studies how public problems are addressed not by single hierarchical agencies but by networks of interdependent organizations — government bodies, nonprofits, firms and community groups — coordinating to deliver services or make policy.
It combines the relational tools of social network analysis with Keith Provan and Patrick Kenis's influential 2008 typology of network governance, which distinguishes shared (participant-governed) networks, lead-organization-governed networks, and network administrative organizations. By mapping the structure of ties, computing network metrics, classifying the governance mode and relating these to outcomes, the method explains how a collaborative network is held together and why it performs as it does.
The intuition behind the method
Imagine the agencies tackling homelessness in a city: a housing authority, several shelters, health clinics, charities and a coordinating office. None controls the others, yet together they form a web of referrals, funding and information-sharing. Network governance analysis treats that web as the object of study, drawing it as a graph of who connects to whom and asking how the web is steered. Is everyone coordinating with everyone (shared governance), does one big agency call the shots (lead organization), or has a dedicated body been set up to run the network (an administrative organization)? Just as a map of roads reveals bottlenecks and hubs, a map of organizational ties reveals which actors are central, how tightly knit the network is, and whether its structure fits the way it is governed.
History
Network analysis of public governance grew out of the policy-network and interorganizational-relations literatures of the 1970s and 1980s and the broader turn from government to governance, in which scholars such as R.A.W. Rhodes argued that policy is increasingly made through self-organizing networks rather than hierarchies. The methodological toolkit came from social network analysis, formalized by Linton Freeman and others. Keith Provan and colleagues advanced whole-network analysis in public administration, and Provan and Kenis's 2008 Journal of Public Administration Research and Theory article 'Modes of Network Governance' provided the typology and contingency theory that anchor much subsequent empirical work on how networks are governed and why they succeed or fail.
When to use it
Use network governance analysis when a public problem is addressed by multiple interdependent organizations and you need to understand how the collaboration is structured, how it is steered, and why it does or does not perform well. It is well suited to collaborative service delivery, interagency coordination, policy networks and partnership evaluation. The method assumes that relational data on the actors and their ties can be collected with adequate coverage, that the network boundary can be sensibly defined, and that the governance-effectiveness contingencies are relevant. It is less appropriate when the system is genuinely hierarchical with a single accountable agency, when relational data cannot be obtained from most actors, or when the question concerns individual organizational performance rather than the network as a whole.
Limitations and considerations
- Network data are demanding to collect; missing actors or unreported ties from low response rates can substantially bias structural metrics.
- Defining the network boundary involves judgment, and different defensible boundaries can yield different conclusions.
- A cross-sectional network snapshot can miss the dynamics of how ties and governance evolve, and effectiveness is hard to attribute to network structure alone.
- The three-mode typology, while influential, simplifies governance arrangements that in practice are often hybrid and shifting.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Makes the invisible structure of interorganizational collaboration explicit and measurable, revealing hubs, brokers and gaps that org charts hide.
- Provan and Kenis's typology links network structure to governance mode and to a clear, testable theory of network effectiveness.
- Combines quantitative network metrics with qualitative governance evidence, supporting both diagnosis and actionable management advice.
- Applicable across many public domains — health, social services, emergency management, environmental governance — where delivery is networked.
Applications
- Analyzing interagency networks delivering health and human services to understand coordination and gaps.
- Mapping emergency-management and disaster-response networks to identify brokers and structural vulnerabilities.
- Studying environmental and natural-resource governance networks spanning agencies, NGOs and communities.
- Evaluating collaborative-governance partnerships and assessing whether their governance mode fits their conditions.
- Examining policy networks to understand how coalitions of organizations shape agenda-setting and implementation.
How it works
How it works
The analyst first defines the boundary of the network — which actors are in scope and what kind of relationship counts as a tie, such as collaboration, referral, funding, information exchange or joint service delivery. Boundary specification is consequential because including or excluding actors and tie types changes every downstream metric; the boundary follows the research or management question, for example all organizations involved in delivering a particular public service in a defined area.
Relational data are gathered on the ties among the actors, typically through surveys in which each organization reports its connections, supplemented by archival records, funding agreements and interviews. The data are organized as a network (an adjacency matrix or edge list) recording who is tied to whom and, where relevant, the direction and strength of each tie. Achieving high response coverage matters greatly, because missing actors or unreported ties distort the picture of the whole network.
Standard social-network metrics summarize the structure: density measures how many of the possible ties actually exist, given here as twice the number of ties divided by the number of possible pairs; centrality measures such as betweenness identify actors who broker connections by lying on many shortest paths between others; and further metrics capture centralization, clustering and the presence of cohesive subgroups. These quantities turn a tangle of relationships into comparable indicators of integration, brokerage and potential vulnerability.
Formula
D = \frac{2L}{N(N-1)}, \qquad C_B(v) = \sum_{s \neq v \neq t} \frac{\sigma_{st}(v)}{\sigma_{st}}Using Provan and Kenis's typology, the network is classified by how it is coordinated: shared governance, in which participants jointly manage the network with no central entity; lead-organization governance, in which one powerful member coordinates the others; or a network administrative organization, a separate entity created specifically to govern the network. The classification draws on the structural metrics and qualitative evidence about decision-making, and it sets up the central question of whether the governance mode fits the network's conditions.
The structure and governance mode are related to network effectiveness, which Provan and Kenis argue depends on the fit between the governance mode and four contingencies: trust among members, the number of participants, goal consensus, and the nature of the task requiring network-level competencies. Dense, high-trust, small networks with strong goal consensus may thrive under shared governance, while larger, lower-trust networks often need a lead organization or administrative organization. The analysis tests this fit against observed outcomes such as service coordination, reach or innovation.
Findings are interpreted to guide network management: whether to invest in building trust, formalize coordination through a network administrative organization, support central brokers, or repair structural gaps that isolate important actors. Because public networks are deliberately steered, the analysis informs concrete choices about governance design and metagovernance rather than merely describing the network, helping managers align structure, governance mode and the demands of the collective task.
Common pitfalls
- Drawing structural conclusions from a network with poor response coverage, where absent actors create artificial gaps and depress centrality and density.
- Treating density or centralization as good or bad in the abstract rather than relative to the network's governance mode and task.
- Forcing a hybrid, evolving arrangement into a single one of the three governance modes and ignoring the mismatch.
- Inferring that a central organization is effective simply because it is structurally central, without evidence on actual network outcomes.
Frequently asked questions
What are the three modes of network governance?
Provan and Kenis identify shared governance, in which all participants jointly manage the network with no central administrative entity; lead-organization governance, in which a single, often more resourceful member coordinates and steers the others; and the network administrative organization, a separate entity created specifically to govern the network. Each mode fits different conditions: shared governance suits small, high-trust, goal-consensual networks, while larger or lower-trust networks tend to require a lead organization or a dedicated administrative organization to be effective.
Why does network boundary specification matter so much?
Every structural metric — density, centrality, centralization — is computed relative to the set of actors included, so deciding who is in the network and what counts as a tie directly determines the results. An overly narrow boundary can make a peripheral broker look central, while an overly broad one can dilute density. Because there is rarely a single objectively correct boundary, analysts justify their choice by the research or management question and often test the sensitivity of conclusions to reasonable alternative boundaries.
How is network effectiveness judged?
Provan and Kenis argue that effectiveness depends on the fit between the governance mode and four contingencies — trust, number of participants, goal consensus, and the need for network-level competencies — and that effectiveness should be assessed at the network level, not just for individual members or clients. In practice analysts pair structural and governance evidence with outcome measures such as service coordination, reach, client outcomes or innovation, recognizing that attributing those outcomes to network structure alone is difficult.
Related methods
Sources
Provan, K. G., & Kenis, P. (2008). Modes of Network Governance: Structure, Management, and Effectiveness. Journal of Public Administration Research and Theory, 18(2), 229–252.
Open DOI referenceCopied.