Regional Environmental Change

, Volume 14, Issue 2, pp 527–539 | Cite as

Policy networks, stakeholder interactions and climate adaptation in the region of South East Queensland, Australia

  • Ryan R. J. McAllister
  • Rod McCrea
  • Mark N. Lubell
Open Access
Original Article


The strategic use of science in regional policy-making forums often assumes collaborative interactions between stakeholders. However, other types of stakeholder interactions are possible. This paper uses the ecology of games to frame an investigation into stakeholder participation in the policy networks for regional climate change planning for South East Queensland, Australia. We tracked organisational participation in policy forums between 2008 and 2012. We then used a novel bipartite network theoretical approach to identify participation by different types of organisations across shared multiple forums, which we argue prefaces: cooperation, collaboration, support or advocacy. Network analysis was then combined with semi-structured interviews to access how scientific information was utilised across the regional network. Our results suggest that stakeholder interactions were predominately used to advocate for organisational agendas. Advocacy artificially narrows the scope of possible policy options and represents a biased, selective use of information. While advocacy is an important part of policy process, as a counter balance, explicit efforts are needed to recurrently expand the scope of policy options.


Social network analysis Exponential random graph model Governance Science and technology policy Science technology and society Political science 


This paper combines quantitative network analysis with qualitative interviews to explore the nature of stakeholder engagement within the regional climate change planning networks of South East Queensland (SEQ), Australia. As in other policy domains featuring high levels of uncertainty, patterns of stakeholder engagement, where stakeholders identify and clarify different policy solutions, play a key role in policy learning. Yet, there is an inherent tension in policy learning between impartiality and stakeholders who strategically advocate for solutions that best serve their interests. This tension occurs because each policy solution distributes benefits and costs in different ways and is heightened by the reality that most policy domains feature multiple decision-making and planning venues in which scientific information may be used.

To analyse the potential for strategic policy learning, Pielke (2007) outlines four modes scientists may adopt when engaging with policy. The ‘pure scientist’ rarely engages with stakeholders and works on problems that are interesting for sciences’ sake, rather than for practical purpose. The ‘arbiter’ engages with key policy stakeholders to help define policy relevant problems, but maintains a linear view of science impact where scientific information is generated independently from policy (Beck 2011). The ‘issue advocate’ strategically uses science to advocate for particular policy solutions, often in cooperation with a smaller coalition of stakeholders with similar policy preferences and values. The fourth type is the ‘honest broker’, who plays a more collaborative role amongst a broad set of stakeholders in the hope of clarifying the scope of policy alternatives and quite possibly increasing the number of alternatives for discussion and associated sense of uncertainty.

Key features distinguishing Pielke’s (2007) different modes of engagement in policy networks are the degree to which stakeholders cooperate with their organisational peers, collaborate across organisations and their level of engagement. Such features define the nature of all stakeholder interactions and not just for those involving scientists. Our hypothesis is that stakeholders predominately engage in policy networks subjectively as advocates, rather than to support policy instruments or to impartially guide open debates on policy options. Exploring the observed mix of the engagement modes in our case study, of which advocacy is one, provides for a constructive analysis on how to manage policy networks with the objective of supporting impartial use of scientific information, thus contributing to debates about the science–policy interface within regional climate adaptation (Hanger et al. 2013).

To frame our data collection, we use the recently revived ‘ecology of games’ which describes patterns of conflict and cooperation that occur in policy domains featuring multiple venues for policy debates (Lubell et al. 2010a). The ecology of games can be operationalised as a policy network, where actors participate in different policy venues. Structural patterns associated with the various types of participating organisations preface the nature of how they interact in the policy network and speak to the ways in which the networks distil scientific information.

In SEQ, the contentious and highly uncertain problem of regional climate adaptation is underpinned by divergent stakeholder preferences (Bohensky and Leitch this edition; Shearer et al. 2013; Taylor et al. in press) and capacities (Keys et al. this edition). How stakeholders engage within the emerging networks determines how science is brought to bear in such policy debates and holds lesson not just in SEQ, but for other regions around the world. During our study period (2008–2012), the state of Queensland was governed by a party which made adaptation and planning for climate change a priority and also supported related scientific research programs (Matthews 2012; Burton and Mustelin in press). While the political landscape changed considerably after elections in March 2012 (Matthews 2012), essentially pausing regional climate change planning, our dataset reflects a period where policy networks developed and operated to mobilise policy, industry and research actors to attempt to understand and resolve regional climate change planning problems.

Policy networks and network configurations

Network approaches to analysing policy systems have focused on what structural conditions matter for different policy and political outcomes, connecting structural characteristics to theoretical concepts. Various case studies have explored how the structure of networks affects governance. For example, based on the case of water management in Western Australia, Robins et al. (2011) linked effective governance to high degrees of network closure. Network closure is the propensity for a group of actors to operate within a tightly bound part, or clique, within a network and is expected to facilitate trust, reputation and cooperation. Sandstrom and Carlsson (2008) likewise suggest that network closure predisposes efficiency. They additionally link heterophily, or a high diversity in actor types, to higher levels of innovation. Based on the case of Swiss Gotthard tourism, Luthe et al. (2012) linked low levels of network density with an inability to collaborate and network centralisation to the capacity to coordinate but an inability to solve unforeseen, complex problems.

What such studies have in common, and what differs from ours, is that they tend (thought not exclusively) to focus on social processes that characterise the entire network. Instead, we are interested in how types of actors participate within networks, exploring various modes of stakeholder interactions in our policy network, and how certain types of organisations are associated with different structural patterns. Other studies have included a sub-network scale focus. Crona and Parker (2011) used network theory to study the utilisation of science in policy using an Arizonan research organisation as its case study. They show that policy makers with more links to a research organisation, and/or to other organisations that utilise science, are themselves more likely to utilise science. Berardo and Scholz’s (2010) network-based ‘risk hypothesis’ discerns between modes of stakeholder interaction, holding that actors who perceive low risks will seek bridging links, which provide access to new information. When the risks are higher, actors will seek more collaborative bonding links, leading to network closure around trusted others. Our study aligns more closely with studies based on how types of actors interact within a network.

Methodologically, our approach focuses on the distribution of the various sub-networks within a policy network (called configurations or motifs). Other authors have also focused on the distribution of configurations (e.g. Lubell et al. 2010b; Robins et al. 2011), and Robins et al. (2012) review the growing body of approaches for robustly examining the prevalence of certain configurations relative to other configurations. As shown by Bodin and Tengo (2012), configurations can be mapped to theory, considerably enhancing the theoretical rigour of network analysis. We build on such studies in a novel way by exploring how selected configurations can then be aligned with modes of stakeholder engagement in policy networks.

Defining and interpreting policy networks

Leading from the ecology of games perspective (Lubell et al. 2010b), we define a policy network as a collection of organisational actors and the forums across which the actors participate in policy debates (i.e. our networks are bipartite). Hence, the actors are not directly linked to other actors, but to the forums within which they choose to participate (Fig. 1). Each organisational actor has attributes—we focus on the type of organisation (local, state, federal government, etc.; see ‘Data and Methods’ section below).
Fig. 1

Stylised policy network with actors (circle) linked to forums (squares) through participation

Table 1 pulls together a nomenclature of the configurations of organisation–forum interactions, which foster different types of stakeholder interactions. The first is a basic configuration (rA) which indicates an organisation’s level of activity (Robins et al. 2011). A given actor can have one link or many. The number of links associated with an actor does not imply anything about the strength of those links. What this speaks to is the breadth (or selectivity) of an organisation’s engagement across the various forums (Lubell et al. 2010b).
Table 1

Network configurations and types of stakeholder interactions, where bold indicates that organisational actors (circles) must be of a given type (e.g. local, state and federal government) (dashed actors can be of any type, squares indicate forums)


Types of stakeholder interactions


Open image in new window

Activity: higher number of links per actor of a specified type of organisation (i.e. fewer links per actor show a organisational type with more limited, more selective engagement)


Open image in new window

Bridging across-type: where one type of organisation disproportionally links to forums attended by different organisational types (given prevalence is relative to TsoA2 in particular). This represents heterophily (interactions between different actor types)


Open image in new window

Bridging within-type: where one type of organisation disproportionally links to forums attended by like organisations (given prevalence is relative to TsoA1 in particular). This represents homophily


Open image in new window

Forum bridging: where one type of organisation disproportionally forms ‘unique’ links between forums. Note that this configuration relates to a single organisational type, so this does not speak to heterophily or homophily of actors interactions


Open image in new window

Closed across-type: where one type of organisation disproportionally forms cliques that include actors with different attributes (given prevalence is relative to C4A2 in particular)


Open image in new window

Closed within-type: where one type of organisation disproportionally forms cliques that include like organisations (given prevalence is relative to C4A1 in particular)

We treat ‘bridging’ configurations (TsoA1, TsoA2 and TsoA) as fostering learning, given these connect actors to others in more distant parts of the network, where information is more likely to be new to them (Granovetter 1973). In these stakeholder interactions, multiple actors are associated with each other via a single, mutually attended forum and can be associated with other actors of the same type (bridging within-type, Table 1, TsA2) or just generically with actors of no particular type (bridging across-type, TsoA1). Where such interactions span organisational types (TsoA1), learning is additionally fostered by ‘heterophily’ or the diversity of actor types (Sandstrom and Carlsson 2008). ‘Bridging’ configurations can also be based on a single organisation that uniquely spans two forums (TsoA).

In contrast, more tightly bonded, ‘closed’ configurations relate to fostering collaborative/cooperative stakeholder interactions, either within-type (Table 1, C4A2) or across-type (C4A1). Within these closed configurations, interactions are more likely to be to others with shared views and knowledge. This fosters shared understandings and expectations, but limits access to new information. Sandstrom and Carlsson (2008) link closure, particularly involving like actors (homophily’), with high levels of cooperation and efficiency in delivering process-based tasks (C4A2). Closure involving different types of organisations prefaces collaboration across organisational types (C4A1).

Exploratory framework and advocacy

A key part of Pielke’s (2007) thesis is that advocacy biases learning and obstructs the efficient use of science for contested problems with uncertain outcomes. The balance of advocacy amongst other modes of stakeholder interactions is a key yardstick in understanding the nature of strategic learning in a policy network. Part of the novelty in our approach is in seeking to identify the balance of advocacy by linking the representation of selected configurations to modes of stakeholder interactions that preface, not just advocacy, but supportive, cooperative, and collaborative interactions as well. A key advantage of such an approach is the potential for greatly improved theoretical power from network theory by mapping to ideas from domains not traditionally adopted by network analysis (Bodin and Tengo 2012). Such a novel approach will require further development, but this research presents its framework as a first step in linking stakeholder interactions with structures found in policy networks.

To develop an exploratory framework based on interpreting network configurations, we start by asking, for any given type of organisation participating in the policy network, which configurations might we expect to see over and/or under represented if they tended to engage as advocates. We build the framework up from Table 1 and present a typology of interaction modes in Table 2, which we suggest preface various types of stakeholder interactions (advocative, collaborative, supportive and cooperative).
Table 2

Interaction modes facilitated by level and type of organisational engagement (↑/↓ refer to greater/lesser representation of configurations than expected by chance)

Advocates will seek to persuade other types of organisations and hence tend to interact more across-type than within-type. Additionally we argue that in seeking to persuade others, they will exhibit bridging relationships as they more frequently interact with stakeholders unlikely to already share the same views and knowledge (bridging across-types, TsoA1↑). We also argue that advocates will have lower, more selective levels of engagement (rA↓). Non-advocates in policy networks for contested problems will need relatively higher levels of engagement in order adequately sift through the broad range of policy solutions. In contrast, advocates come with a pre-defined policy solution and so engage less given their more simplified rationale for participation.

To complement our exploration of advocacy, we further define pairs of configurations as stakeholder interactions prefacing supportive, cooperative and collaborative modes of engagement.

Stakeholders taking a ‘supportive’ role we define as likewise engaging with a small, targeted set of stakeholders (rA↓) but they disproportionally interact within-type in supporting their organisational peers (TsoA2↑). We define ‘cooperative’ and ‘collaborative’ stakeholder interactions as organisations which engage with a broader set of stakeholders (rA↑). ‘Cooperative’ organisations are tightly bonded in closed configurations with the same types of organisations (C2A2↑). In contrast, ‘collaborative’ organisations operate in closed configurations across various types of organisations.

There may also be other modes of stakeholder interactions. However, this research presents its framework as a first step and builds on previous research by linking stakeholder interactions with constellations of network configurations. By exploring whether such patterns of configurations are more or less frequent in a policy network than expected by chance and combining this with information from stakeholder interviews, statements can be made as to which organisational types engage in different modes of stakeholder interaction. This builds toward more concrete thinking about what we mean when we relate structure to network function, which is somewhat missing in the literature.

Data and methods

Network data

Network data were collected iteratively. First, attendance records for key, known forums were obtained. To be in scope, a forum had to have a focus on climate change planning or adaptation and with a strong (though not exclusive) focus on the region of SEQ (Table 3). It must also have sponsored debate and negotiation, rather than one-way communication, so as to facilitate stakeholder interactions. Our sample period covers 2008 to mid-2012. Network data and our list of forums were checked with participants as part of our semi-structured interviews. Participants were also asked to suggest additional forums with potential significance to climate planning in SEQ.1
Table 3

Forums relating to climate change planning and adaptation in South East Queensland (SEQ)


Time frame/acronym

Meeting dates

Representation (attendance)

(1) Climate Change Management Plan—Working Group 1—Regional Climate Projections and Impact Modelling

2008 WG1

28 May 2008; 16 June 2008; 30 June 2008

Attendance records

(2) Climate Change Management Plan—Working Group 2—Regional Vulnerability Assessment and Adaptation Planning,

2008 WG2

28 May 2008; 16 June 2008

Meeting notes/attended by lead author

(3) Climate Change Management Plan—Working Group 3—Regional Greenhouse Gas Investigations and Mitigation Planning

2008 WG3

28 May 2008; 16 June 2008; 30 June 2008

Attendance records

(4) Climate Change Management Plan—Steering committee—oversee working groups

2008 SC

26 March 2008; 30 April 2008; 25 June 2008; 30 July 2008; 03 September 2008; 29 October 2008; 17 December 2008

Meeting notes/attendance records/attended by lead author

(5) Climate Change Management Plan—Terms of Reference Committee

2008 TOR

07 April 2008

Meeting notes

(6) Climate Change Management Plan—Regional Planning Information Inventory Gap Analysis

2008 WS1

27 February 2008

Meeting notes

(7) Climate Change Management Plan—Priorities for Adaptation Workshop

2008 WS2

14 April 2008

Meeting notes

(8) SEQ Urban Water Security Research Alliance—Climate and Water Project Reference Group Meeting

2008–2012 UWSRASC

8 March 2008; 3 November 2009; 13 April 2010; 7 September 2010; 9 August 2011; 24 February 2012

Meeting notes/attended by lead author

(9) SEQ Climate Adaptation Research Initiative—Steering Committee

2009–2012 CARISC

20 November 2009; 21 May 2010; 22 June 2010; 2 March 2011; 5 September 2011

Meeting notes/attended by lead author

(10) Climate Q—Helping Primary producers adapt to Climate change—Steering committee


24 August 2010; 18 October 2010; 15 July 2011; 28 October 2011

Meeting notes

(11) SEQ Climate Adaptation Research Initiative—Human Settlements Project Reference Group

2010–2012 PRG

10 June 2012; 8 November 2012; 5 September 2011

Meeting notes/attended by lead author

(12) SEQ Climate Adaptation Research Initiative—Human Settlements Project Workshop

2011 PRGWS

17 November 2011


(13) SEQ Adaptation Research Liaison Group

2009 LI

9 April 2009; 28 July 2009; 29 September 2009

Attendance records

(14) Inland Flooding Scientific Advisory Group meetinga

2010 INFSC

4 March 2010; 12 July 2010; 23 August 2010

Attended by lead author

(15) Greencross Hardenup—Project Reference Committeea

2011 GCHU

16 May 2011;15 July 2011;8 August 2011

Attendance records

(16) Greencross King Tides—Steering Committeea

2011–2012 GCSC

10 January 2012; 28 February 2012

Meeting Notes

aCoded as directly supporting an existing policy instrument

Three of the 16 forums’ attributes were coded as having a clear defined role in terms of supporting a specific ‘policy’ instrument (Table 3). Other forums frequently had fairly broad agendas and generally aimed at some form of consensus rather than providing either data or decisions. Organisational attributes were coded by type, as either local, state or federal government, or consulting, research, industry or non-government organisation (Table 4).
Table 4

Organisational actors who participated in Table 3’s forums relating to climate change planning and adaptation in South East Queensland (SEQ). Subheadings indicate the type of organisation

Policy actor (organisational level)


Non-government sector


 Queensland Council of Social Service


 Non-government Individual






 Climate Risk Pty Ltd


 Mercury Project Solutions


 Maunsell AECOM


 DHI Consultants


Australian government sector


 Department of Climate Change and Energy Efficiency


 Bureau of Meteorology


Industry and representative groups


 DERM—Queensland Water Commission


 Queensland Farmers Federation




 SEQ Catchments


Local councils (government) and representative groups


 Western Suburbs Regional Organisation of Councils


 Toowoomba Regional Council


 Sunshine Coast Regional Council


 Redland City Council


 Queensland Coastal Council Group


 Moreton Bay Regional Council


 Lockyer Valley Regional Council


 Local Government Association of Queensland


 Logan City Council


 Ipswich City Council


 Gold Coast City Council


 Council of Mayors (SEQ)


 Brisbane City Council




 Emergency Management Aust. Institute Centre for Excellence


 Walker Institute for Climate System Research


 Urban Water Security Research Alliance


 University of Queensland


 SEQ Healthy Waterways Partnership


 Queensland University of Technology


 National Climate Change Adaptation Research Facility


 Griffith University


 Commonwealth Scientific and Industrial Research Organisation


State Government and affiliated


 Urban Development Institute of Australia


 Queensland Health


 NSW Environment—Climate Change and Water


 Department of Transport and Main Roads—Transport


 Department of Transport and Main Roads—Main Roads


 Department of the Premier and Cabinet


 Department of Communities


 Department of Local Government and Planning


 DERM—Water and Water Policy


 DERM—Queensland Climate Change Centre for Excellence


 DERM—Office of Climate Change


 DERM—Mines and Energy




 DEEDI Primary industries and fisheries


 DEEDI—Economic Development (tourism, regional)


 EMQ—Emergency Management Queensland


 DCS—Department of Community Safely


The attendance records showed individuals, while our interest was in organisational representation. Hence, we coded attendance as the organisation which the individual represented at the time of forum sitting. To account for Queensland Government departmental restructuring in 2009, our organisational coding was based on functional areas, based largely on 2009 departmental titles. This allowed us to code consistently across restructures. For example, the ‘Environmental Protection Agency’ (EPA) was abolished in 2009 and their activities moved to the new the ‘Department of Environment and Resource Management’ (DERM). EPA activities were coded as ‘DERM—Environment’. Post-2009, DERM participants focusing on environmental management were also coded as ‘DERM—Environment’ and most of these functions had in fact previously been housed within the EPA.

Figure 1 depicts a stylised network, while Fig. 2 depicts the actual network (raw data available in electronic supplement 1). Our network formulation follows Lubell et al. (2010b). Organisations are linked to any forum which had participation by an individual while representing that organisation. The network links are not weighted, meaning that a link based on an actor attending a selected forum once was treated equally to where an actor attended a forum many times. Returning to our typology for identifying modes of stakeholder interactions (see Table 2), if an organisation demonstrates a greater number of rA configurations, than by chance this allows interpretation about the breadth of stakeholder engagement, but not the strength/depth. Similarly, all forums are treated equally, regardless of whether a forum sat once or several times. Hence, the power of network theory is in rather clinical explanations of structure alone. We complement such analyses with qualitative data, which allows us to unpack the processes underpinning the structural aspects of the data (below).
Fig. 2

Graphical representation of network data. Forums are depicted as squares (Table 3)—organisations as circles (Table 4). Organisations have sizes scaled by betweenness centrality (see Figure A1, Supplementary material 2)

Network configurations

Exponential Random Graph Modelling (ERGM) is used to describe the relative prevalence of particular configurations (e.g. Table 1) within networks (see Robins and Morris 2007; Robins et al. 2007a, b). ERGMs are essentially logistic regressions where the dependant variables are the instances where a link exists between two nodes, and the configurations in the network are the explanatory data (i.e. predicting whether an organisation attended a forum based on their position in the network). ERGMs assume network links are part of a stochastic process. This means they assume an observed network is just one expression of some underlying process, allowing statistical tests without the need for comparative networks. By estimating what might be expected by chance specifically given the distribution of other configurations minimises the potential for mis-interpretation (Robins et al. 2012). Other approaches, such as stochastic-actor models (Snijders 2001) and relational events models (Butts 2008), explore longitudinal datasets. While our forums occur over time, there is considerable overlap in timing and uneven longevity of the forums (and organisational participation). Hence, we analyse our network as a single dataset. We used the computer package pNet (bpNet for bipartite networks, Wang et al. 2009).

Semi-structured interviews

We used semi-structured interviews to complement our network analysis. We sourced participants who had been more active in the policy network. Preliminary network analysis showed that state and local governments had both high levels of participation in forums and were both associated with significant explanatory power (i.e. significant configurations in the ERGM, see below). Hence, we interviewed key state departments, and local government and their representatives, who we knew to be key information users. Five interview participants came from state government, eight from local government (representing six different regional councils), and two from local government representative bodies. Interviews lasted between 40 and 60 min. Questions centred on why and how organisations participated and how they used scientific information, including
  • What would a good outcome for your organisation look like as a result of participating in these forums?

  • How does your organisation’s perspective compare with others at these forums?

  • Within these forums, what scientific information did you find most useful? When, where and in what formats? How were they useful?

  • What types of scientific information did you find least useful (and in what types of formats?)

  • What scientific information is missing that would be useful?

  • How have you used scientific information in achieving your organisational objectives?

  • How did other organisations use science information at these forums?

These interviews were recorded, transcribed, de-identified and coded. The text data were coded using NVivo into broad categories relating to the above questions. Common themes within each broad category emerged and selected text was then coded into sub-categories, which often related to organizational types. This provided qualitative data on organizational motivations, agendas, uses of science information, opportunities and challenges, as well as formal and informal interactions inside and outside the policy network.

Quantitative results

Exponential Random Graph Model (ERGM)

Our ERGM includes 15 configurations, which collectively explain the network’s overall structure (Table 5, and also see Supplementary material 2 for further modelling detail). Ten of these are versions of configurations described above (Table 1), which relate to specific types of organisations. Five more are more general configurations (Fig. 3.), which are included to control for the influence of the overall structure of the network.
Table 5

ERGM model: estimated parameters and observed configuration counts for an Exponential Random Graph Model with a fixed density = 0.25481: */**/*** showing 90/95/99 % significance for the parameters


Parameter estimates (t stat)

Observed counts (t stata)


0.0091 (4.5465)***

18612 (−0.02)


−1.3709 (−3.2562)***

259 (−0.02)


−2.9396 (−0.9606)

361 (0.02)


0.3049 (1.6480)*

192 (0.01)


−0.1387 (−4.0797)***

1039 (−0.01)

rA (State Gvt)

−0.3901 (−1.2747)

86 (0.04)

rA (Local Gvt)

−1.3507 (−2.9883)***

45 (0.00)

rA (Consultancy)

−2.2856 (−2.1163)**

6 (0.01)

TsoA (Research)

0.0467 (1.3341)

99 (−0.04)

TsoA1 (Local Gvt)

0.1841 (2.7899)***

387 (−0.03)

TsoA1 (Consultancy)

0.1170 (2.3398)**

161 (0.00)

TsoA2 (State Gvt)

0.1308 (3.1902)***

357 (0.03)

TsoA2 (Local Gvt)

−0.2791 (−1.5336)

115 (−0.02)

c4A2 (Local Gvt)

0.0641 (1.8861)*

167 (0.03)

TsoP1(Instrum. forum)

−0.0968 (−1.6975)*

93 (−0.01)

a T tests compare observed configuration counts against simulation means

Fig. 3

Generic configurations included used in Exponential Random Graph Model. Circles indicate organisations of no particular type; squares forums. See Wang et al. (2009) for more information

The model shows that 11 configurations are observed more frequently than expected by chance, given the distribution of other network patterns. Having significantly fewer stars (K-Sa) suggests a low propensity to cluster around key organisations. L3 configurations are over represented. These are essentially loose threads on the periphery of the network. To have more of these, given the abundance of configurations like K-Cp and K-Ca, implies that the network also contains regions of relative density and clustering (i.e. where the L3s are not). Both K-Ca and K-Cp relate to closure, but the first is significantly positive and the second negative: there is closure in the network, but only focused around key forums, not key organisations.

While our model controls for these broarder aspects of network function, our main interest relates to how different types of organisations interact within the network. Hence, key to our discussion is the results that relate to the over and/or under representation of configurations for the various types of organisations:
  • Consulting organisations tended to have significantly more cross-type bridging configurations (TsoA1) and less activity (rA) than what could be expected given its representation in the policy network. According to our typology, these collectively imply an issues ‘advocacy’ role (Table 2).

  • Local governments had significantly greater representations of closed within-type configurations (C4A2) and across-type bridging (TsoA1) configurations, but significantly low numbers of rA which pointed to lower level of, and selective, engagement. The significance (and signs) of TsoA1 and rA implies local governments take on an ‘advocacy’ role, but in addition to significant closed within-type configurations (C4A2).

  • State government demonstrated a statistical propensity to have more within-type bridging stakeholder interactions (TsoA2). The activity parameter (rA) was positive, but while deemed important enough to retain in the model (see Supplementary material) was not significant.

  • No significance was found for federal government, research, NGO or industry configurations. This may reflect their limited number of participating organisations.

Qualitative data

The qualitative data are useful for exploring the different uses of science information by stakeholders and how this interacts with different modes of stakeholder interaction. As mentioned in the methods in ‘Data and methods’ section, two predominant types of stakeholders were followed up with qualitative interviews: local and state government stakeholders.

The local government officers generally sought science information to understand the complexity of climate change, especially any implications for small spatial areas, to facilitate planning in their local government areas.

Like it wasn’t brought down to the local government level or even to SEQ [region] level, a lot of the information. So it was when you’re talking to engineers or to planners they want more detailed information. (Local Gvt, officer 73)

This desire of local government officers to understand the implications of climate change for their local areas was balanced by a need to communicate climate change implications simply and clearly to their councillors to gain local political support for adaptation plans. This suggests local government officers are also issue advocates outside the regional policy network.

We had to bring them along on the journey and get their understanding quite clear about climate change. Because obviously some were climate change sceptics and we had to sort of answer their questions, so it was about getting really clear messages and clear slides.… So, it’s not like they’ve got a huge background in it at all. (Local Gvt, officer 73)

State government organisations faced a similar situation where different state organisations worked with different levels of climate change complexity. For example, state government research organisations produce detailed climate change information, while state government policy organisations often preferred less detailed climate change information.

So the government just sticking to an A1FI policy as opposed to looking at a range of emission scenarios… It happens, so the trick as a scientist is to know it’s going to happen and to try and incorporate it in a way that’s digestible, which is what we try to do with our advice. We try to couch it in such a way that we can still get that risk profile in there somehow. (State Gvt – scientist)


Policy Officer 17: … it will get hotter and warmer, dryer, wetter, whatever. For a policy person they’re useful words that you can start feeding up. I personally don’t need to go into the detail of the data… You’re also very mindful that you’ve got a very credible source of information. (State Gvt, policy officer 17).

Thus, science information was used in a variety of ways by stakeholders in perusing their own agendas, often adopting advocative modes of interaction when engaging across organisational types.
Within-type cooperative interactions were more evident between environmental officers in local government organisations.

Well we just run into each other at different adaptation meetings, well relating to adaptation and always discuss how our councils are going and how we’re implementing it and what challenges are we having. So that’s an informal network. (Local Gvt, officer 73)

This provides some qualitative support for the risk hypothesis (see Berardo and Scholz 2010) where local government cooperates more because of their increased vulnerability to climate change impacts. As well as participating in these forums, this group of local government environmental officers also met regularly outside of these forums, though less formally, to discuss similar issues. So the forums capitalised upon and further facilitated common understandings, knowledge and cooperation at the local government level.
Collaborative interactions across types of organisations were complicated by different levels of prior knowledge about climate change which suggests that forums may be better managed to facilitate collaboration and learning.

I was frustrated because … the same questions were being asked through lack of knowledge, lack of understanding, which had been going on for twenty years. I just got the feeling we hadn’t advanced. (Local Gvt, officer 73)

Actors had different levels of understanding about climate change issues and different needs for different levels of understanding. For example, the need was for more complex understandings for state government research organisations and local government planning, and more simplified understandings (with credible sources) for policy and political purposes. In another example, cross-type collaboration between local government and other levels of government was complicated by needs for different geographic levels of information (e.g. small area vs. regional information).
Across-type collaboration between local government and other levels of government was also complicated by existing power relationships which meant that local governments were primarily feeding up information to state government bodies for their draft regional plans as a way of advocating for their own needs in the process.

Probably with some of them it was more like the state dragging information out of - or taking information from local governments and then nothing coming back. But I mean, if you didn’t give that information then you wouldn’t have any say. (Local Gvt, officer 73)

Similarly, collaboration can be complicated by power relations between state government organisations and ministers such that regional plans do not necessarily get approved without support at the highest levels of government.

… it comes down to basically internal government processes and whose minister has got more clout than someone’s minister. So the planning minister generally has more clout than the environment minister. So there were serious loggerhead discussions about that plan and what it could and couldn’t do … Ultimately that’s why, I think, it went nowhere. (State Gvt, policy officer 288)

Clearly, collaboration across organisational types cannot be assumed in policy networks. Given the different agendas, needs and power relations between different types of stakeholders, advocacy is a more likely mode of interaction across stakeholder types. Any attempts to facilitate collaboration need to be carefully managed.
In a similar vein, just participating in forums did not mean that organisations necessarily engaged in meaningful stakeholder interactions. Some participants felt that forum activities focused primarily on getting people together as the actual outcome, as opposed to any subsequent decision making.

… I go along to some of these sessions and an achievement will seem to be that all the important people are involved [but that] doesn’t mean that they’ve agreed or signed up or are strong champions or promoters…. I know good efforts were made to try and make sure that stakeholders were informed and engaged, but I’m not sure that actually that was achieved. (State Gvt, policy officer 16)

Our qualitative data suggest the most common modes of stakeholder interaction were within-type cooperation and across-type advocacy. It seems as if within-type cooperation can displace within-type support and across-type advocacy can displace across-type collaboration.

Different actors and organisations within the network have different roles, levels of understanding and different information needs to perform their roles within the network. Thus, the efficient use of science information in the network needs to involve various modes of stakeholder interaction using science information pitched at the needs and roles of the different actors.

Discussion and conclusions

Identifying modes of stakeholder interactions in networks

Local government showed three significant network configurations, which when combined with qualitative data allow us to tell a rich narrative around both cooperation and issues advocacy. The qualitative data suggested that local governments cooperate to share important information and to discuss agendas. Our quantitative results also suggested within-type cooperation (C4A2, Table 5). On the other hand, local governments also demonstrate characteristics of advocacy in the qualitative and quantitative data, with fewer links to other network stakeholders (rA-local govt) and positive across-type stakeholder interactions (TsoA1). Overall, our analysis suggests local governments cooperate as a closed group to form a shared agenda and then lobby other organisational types in an advocacy role.

In contrast to local government actors, the state government showed more within-type supportive stakeholder interactions (TsoA2) according to the quantitative data. The qualitative data, however, strongly questions the degree to which various state organisations share values and agendas and hence the degree to which they may be considered within-type (e.g. departments responsible for the environment had different agendas to departments responsible for infrastructure, and departments responsible for climate change research had different information needs to those responsible for policy). Thus, within-type stakeholder interactions may misrepresent interactions for state organisations, and care should be taken not to code within-type interactions too broadly. It seems that for state organisations, within-type stakeholder interactions were more like across-type stakeholder interactions. Noting too that the activity parameter was negative, but insignificant (see rA-local, Table 5). We would argue therefore that on the balance of qualitative and quantitative evidence, state organisations seem most engaged in advocacy stakeholder interactions.

Consultant organisations demonstrated fewer links to other network actors than expected by chance (see rA-consultant, Table 5) and more across-type stakeholder interactions (TsoA1). This accords with advocative stakeholder interactions, where such organisations engage with a particular agenda.

Overall, our data suggest that organisations within our SEQ policy network were most engaged as ‘advocates’. Consultants and local government both engaged in stakeholder interactions that we define as prefacing ‘advocacy’, and there is an argument that state organisations did the same. Our analysis also points to the multifaceted aspect of participants, with local governments both advocating and cooperating where appropriate. However, there was little evidence of collaborative stakeholder interactions across both qualitative and quantitative data.

Managing policy networks for effective learning

Formal governance structures of checks and balances can be designed and implemented. Policy networks on the other hand can only be managed (Klijn et al. 1995; Klijn and Koppenjan 2000, 2006; van Bueren et al. 2003). The objective of managing policy networks is not in achieving a stated outcome necessarily but rather in promoting the effectiveness of the network to facilitate the various embedded policy ‘games’. Given advocative stakeholder interactions seem to dominate the policy network, and thinking specifically about learning and using scientific information (regardless of the policy outcome), we may ask, what would be different if our network were better managed to ensure efficient use of science? And how can we achieve this?

Beck (2011) argued that the linear model of science has prematurely narrowed the scope of policy debates on adaptation. And Hanger et al. (2013) likewise identify a need for richer stakeholder engagement in climate debates in order to improve the science–policy interface. Pielke’s (2007) ‘honest broker’ mode of how scientists could engage in policy also talks to these issues. Where the outcomes from various policy actions are highly uncertain, ‘honest brokers’ are critical because they, by seeking to expand problem scope, work to offset the narrowing role of ‘issue advocates’. Advocacy and lobbing are critical ingredients for achieving policy decisions in contested spaces. However, the biased learning associated with advocacy does not yield efficient use of science for contested problems with uncertain outcomes.

In our more general typology, the honest broker role accords most closely with collaborative interaction modes involving research organisations. Research organisations and federal government are perhaps most likely to play an enhanced role in broadening the scope of policy alternatives in collaborative engagement with other stakeholders. In practice, however, this would involve increasing and broadening engagement with stakeholders, which will require new, targeted funding, initiatives or incentives.

In contrast, collaboration is perhaps not a role that can be expected of state or local government because they do and should have strong local agendas, and as such advocacy is a legitimate role. Neither is it the role of NGOs, industry or consultants. However, the balance between various stakeholder interaction modes can be managed by better structuring engagement so that advocacy can be channelled so that it becomes more functional, for example, by sharing power between levels of government, encouraging collaborative interactions and involving unaligned stakeholders such as research organisations.

Wagenaar (2011) reviews collaborative policy making, where empowered stakeholders contest in open-minded, reciprocal debates. In such debates, knowledge’s worth is judged in relationship to its practical solutions. Our qualitative data point to the diversity in expectations and requirements of various stakeholders. The ‘practical solution’ for an advocate may simply be an achieved political agenda, whereas for a local government policy officer, it may be on-ground action. The style and complexity of information requirements are equally diverse. Our discussions with stakeholders also pointed to the frustrations where those seeking learning were mixed in the one forum with those seeking decisions, for example. There needs to be greater clarity within discrete forums such that stakeholders can be meaningfully empowered to deliver. In tightening the roles/agenda of forums, additional forums will likely be needed to accommodate the existing diversity of objectives.

Efforts to improve policy network efficiency need to reflect not just how to intervene, but when. Berardo and Scholz’s (2010) risk hypothesis holds that new policy networks, where the risk of stakeholder ‘defection’ is low, stakeholders will favour bridging links, which we interpret more broadly to include across-type stakeholder interactions. As policy networks mature, stakeholders refine the problems that they are seeking to resolve. Correspondingly, as attention turns toward action, the risks associated with making more binding decisions increase, and now, stakeholders will prefer bonding links, i.e., within-type cooperative stakeholder interactions. For example, a study of regional flooding adaptation in the Netherlands argued that the related networks had collaborative linkages between stakeholders during master planning and more isolated interactions during implementation (van Buuren et al. in press). In Australia, local government is at the ‘coal-face’ of action on climate adaptation and for them the potential costs of poor outcomes are highest (Barnett et al. 2013; Fletcher et al. 2013). Correspondingly, local government was also the only organisational type with a significant representation of within-type cooperative stakeholder interactions (Table 5, C4A2). While our data do not allow us to explore Berardo and Scholz’s (2010) risk hypothesis temporally, our results suggest it may hold across space. The implication is that given cooperation already exists in the parts of the network facing the greatest risks, then the need for intervention and more structured engagement will be in the parts of the networks where the risks are actually lowest.

Data on climate change planning in SEQ suggest policy networks are not simply a means for fostering collaboration. Instead, they are an emergent structure through which contestation, vested interests, factions and cooperatives are partitioned through selective participation in various debates and sub-issues. Our results suggest that within these policy networks for uncertain, contested problems, advocacy is the dominate mode of stakeholder interaction, and collaboration cannot be assumed. In order to counter balance advocacy, rather than trying to simplify stakeholder engagement, we need to facilitate more diversity in engagement with greater clarity in purpose.


  1. 1.

    The lead author had extensive prior knowledge, having represented the research organisation ‘CSIRO’ in six of the studied 16 forums in various capacities, and no new forums were identified as part of this cross-checking and snow-balling process.



This research is part of the South East Queensland Climate Adaptation Research Initiative, a partnership between the Queensland and Australian Governments, the CSIRO Climate Adaptation National Research Flagship, Griffith University, University of the Sunshine Coast and University of Queensland. We received ethics clearance for both the network data collection (31 January 2011) and semi-structured interviews (22 April 2011) through CSIRO’s Social Science committee (#006/12). Many thanks to Lisa Brennan, Catherine Lovelock, Liana Williams, Brian Coffey, two anonymous reviewers and the many seminar attendees who provided us with critical comments.

Supplementary material

10113_2013_489_MOESM1_ESM.docx (25 kb)
Supplementary material 1 (DOCX 25 kb)
10113_2013_489_MOESM2_ESM.docx (44 kb)
Supplementary material 2 (DOCX 45 kb)


  1. Barnett J, Waters E, Pendergast S, Puleston A (2013) Barriers to adaptation to sealevel rise. National Climate Change Adaptation Research Facility, Gold CoastGoogle Scholar
  2. Beck S (2011) Moving beyond the linear model of expertise? IPCC and the test of adaptation. Reg Environ Change 11:297–306. doi: 10.1007/s10113-010-0136-2 CrossRefGoogle Scholar
  3. Berardo R, Scholz JT (2010) Self-organizing policy networks: risk, partner selection, and cooperation in estuaries. Am J Polit Sci 54:632–649CrossRefGoogle Scholar
  4. Bodin O, Tengo M (2012) Disentangling intangible social-ecological systems. Glob Environ Change 22:430–439. doi: 10.1016/j.gloenvcha.2012.01.005 CrossRefGoogle Scholar
  5. Bohensky E, Leitch A (this edition) Framing the flood: a media analysis of themes of resilience in the 2011 Brisbane flood. Reg Environ Change doi: 10.1007/s10113-013-0438-2
  6. Burton P, Mustelin J (in press) Planning for climate change: is greater public participation the key to success? Urban Policy Res. doi: 10.1080/08111146.2013.778196
  7. Butts CT (2008) A relational event framework for social action. Sociol Methodol 38:155–200Google Scholar
  8. Crona BI, Parker JN (2011) Network determinants of knowledge utilization: preliminary lessons from a boundary organization. Sci Commun 33:448–471. doi: 10.1177/1075547011408116 CrossRefGoogle Scholar
  9. Fletcher CS, Taylor BM, Rambaldi AN, Harman BP, Heyenga S, Ganegodage K, Lipkin F, McAllister RRJ (2013) Costs and coasts: an empirical assessment of physical and institutional climate adaptation pathways. National Climate Change Adaptation Research Facility, Gold CoastGoogle Scholar
  10. Freeman LC (1977) A set of measures of centrality based on betweenness. Sociometry 40:35–41CrossRefGoogle Scholar
  11. Granovetter MS (1973) The strength of weak ties. Am J Sociol 78:1360–1380CrossRefGoogle Scholar
  12. Hanger S, Pfenninger S, Dreyfus M, Patt A (2013) Knowledge and information needs of adaptation policy-makers: a European study. Reg Environ Change 13:91–101. doi: 10.1007/s10113-012-0317-2 CrossRefGoogle Scholar
  13. Keys N, Bussey M, Thomsen DC, Lynam T, Smith TF (this edition) Building adaptive capacity in south east Queensland, Australia. Reg Environ Change doi: 10.1007/s10113-012-0394-2
  14. Klijn EH, Koppenjan JFM (2000) Public management and policy networks. Publ Manag 2:135–158Google Scholar
  15. Klijn EH, Koppenjan JFM (2006) Institutional design—changing institutional features of networks. Public Manag Rev 8:141–160. doi: 10.1080/14719030500518915 CrossRefGoogle Scholar
  16. Klijn EH, Koppenjan J, Termeer C (1995) Managing networks in the public sector- a theoretical study-of management strategies in policy networks. Public Admin 73:437–454. doi: 10.1111/j.1467-9299.1995.tb00837.x CrossRefGoogle Scholar
  17. Lubell M, Henry AD, McCoy M (2010a) Collaborative Institutions in an ecology of games. Am J Polit Sci 54:287–300. doi: 10.1111/j.1540-5907.2010.00431.x CrossRefGoogle Scholar
  18. Lubell M, Robins G, Wang P (2010b) Policy coordination in an ecology of water management games, Annual meeting of the midwest political science association, ChicagoGoogle Scholar
  19. Luthe T, Wyss R, Schuckert M (2012) Network governance and regional resilience to climate change: empirical evidence from mountain tourism communities in the Swiss Gotthard region. Reg Environ Change 12:839–854. doi: 10.1007/s10113-012-0294-5 CrossRefGoogle Scholar
  20. Matthews T (2012) Responding to climate change as a transformative stressor through metroregional planning. Local Environ 17:1089–1103. doi: 10.1080/13549839.2012.714764 CrossRefGoogle Scholar
  21. McAllister RRJ, Izquierdo LR, Janssen MA, Stafford Smith DM (2009) Research impact within the international arid literature: an Australian perspective based on network theory. J Arid Environ 73:862–871. doi: 10.1016/j.jaridenv.2009.03.014 CrossRefGoogle Scholar
  22. Pielke RA (2007) The honest broker: making sense of science in policy and politics. Cambridge University Press, CambridgeCrossRefGoogle Scholar
  23. Robins G, Morris M (2007) Advances in exponential random graph (p*) models. Soc Netw 29:169–172. doi: 10.1016/j.socnet.2006.08.004 CrossRefGoogle Scholar
  24. Robins G, Pattison P, Kalish Y, Lusher D (2007a) An introduction to exponential random graph (p*) models for social networks. Soc Netw 29:173–191. doi: 10.1016/j.socnet.2006.08.002 CrossRefGoogle Scholar
  25. Robins G, Snijders T, Wang P, Handcock M, Pattison P (2007b) Recent developments in exponential random graph (p*) models for social networks. Soc Netw 29:192–215. doi: 10.1016/j.socnet.2006.08.003 CrossRefGoogle Scholar
  26. Robins G, Bates L, Pattison P (2011) Network governance and environmental management: conflict and cooperation. Public Admin 89:1293–1313. doi: 10.1111/j.1467-9299.2010.01884.x CrossRefGoogle Scholar
  27. Robins G, Lewis JM, Wang P (2012) Statistical network analysis for analyzing policy networks. Policy Stud J 40:375–401CrossRefGoogle Scholar
  28. Sandstrom A, Carlsson L (2008) The performance of policy networks: the relation between network structure and performance. Policy Stud J 36:497–524. doi: 10.1111/j.1541-0072.2008.00281.x CrossRefGoogle Scholar
  29. Shearer H, Taygfeld P, Coiacetto E, Dodson J, Banhalmi-Zakar Z (2013) The capacities of private developers in urban climate change adaptation (Final report), National climate change adaptation research facility, Gold CoastGoogle Scholar
  30. Snijders TAB (2001) The statistical evaluation of social network dynamics. Sociol Methodol 31:361–395. doi:  10.1111/0081-1750.00099 Google Scholar
  31. Taylor BM, Wallington TJ, Heyenga S, Harman BP (in press) Urban growth and climate adaptation: Divergent discourses and implications for governance. Urban Stud. doi: 10.1177/0042098013484529
  32. van Bueren EM, Klijn EH, Koppenjan JFM (2003) Dealing with wicked problems in networks: analyzing an environmental debate from a network perspective. J Publ Adm Res Theor 13:193–212. doi: 10.1093/jopart/mug017 CrossRefGoogle Scholar
  33. van Buuren A, Driessen P, Teisman G, van Rijswick M (in press) Toward legitimate governance strategies for climate adaptation in the Netherlands: combining insights from a legal, planning, and network perspective. Reg Environ Change doi: 10.1007/s10113-013-0448-0
  34. Wagenaar H (2011) Meaning in action: interpretation and dialogue in policy analysis. M.E. Sharp, New YorkGoogle Scholar
  35. Wang P, Robins G, Pattison P (2009) Pnet—program for the simulation and estimation of exponential random graph (P*) models. University of Melbourne, MelbourneGoogle Scholar

Copyright information

© The Author(s) 2013

Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.

Authors and Affiliations

  • Ryan R. J. McAllister
    • 1
  • Rod McCrea
    • 1
  • Mark N. Lubell
    • 2
  1. 1.CSIRO Ecosystem SciencesBrisbaneAustralia
  2. 2.University of California DavisDavisUSA

Personalised recommendations