Hamalt genetics based peer to-peer network architecture to encourage the coo...
Master thesis presentation
1. Complex climate change adaptation network in Bhutan;
how actor type impacts actor inclusion and clustering
Master’s Thesis
Institute of Political Science
University of Bern, Switzerland
Presented by:
Jessica Russell
17 September 2015
Thesis supervisor:
Prof. Dr. Karin Ingold
Institute of Political Science
University of Bern, Switzerland
2. Background
> Policy processes and governance as networks:
Actors participate in collective decision-making process’ called an “institution”
Repeated transactions between actors become institutionalized in a “network”
Actors collaborate within the network on a defined project, creating a “tie”
Stakeholder inclusion, scales and multi-level approaches:
Climate change adaptation is local in scope, with the structure of institutional
arrangements being critical, local actor inclusion is important
The multilevel governance framework involves two levels of action:
1) Vertical = crossing multiple government scales - local to national
2) Horizontal = across government departments, line ministries
Vertical/horizontal integration through top-down/bottom-up institutional designs
A “polycentric approach” to governance
Case study: Bhutan
3. Research Question 1
> Aim: Examine how actor inclusion in climate change adaptation institutions is
shaped in a complex network in Bhutan, by applying a multilevel governance
framework to explore links between actors and institutions within networks
> Research Question: Which type of actors are most included within complex
networks and projects in climate change adaptation policy in Bhutan via the
embededness of national and local actors in top-down and bottom-up institutional
designs
> Hypothesis: National actors are well embedded in top-down institutional designs,
and local actors are well embedded in bottom-up institutional designs
Independent Variable: Actor type
Dependent Variable: Actors’ inclusion
4. Data and Methodology – Research
Questions 1, 2, 3
> 130 key actors were identified from a literature review and defined as nodes,
categorised into 7 actor types: international organisation, national government,
foreign government, local government, NGO, corporation or community
> 73 climate change adaptation-related policies/projects defined and categorised
into:
Institutional types: climate change adaptation, sustainable forestry, disaster
management, communication (14)
Institutional designs: top-down (69) or bottom-up (4)
Social Network Analysis (SNA). Software- UCINET:
Ties between actors and institutions coded as a dummy variable, with 1: tie
being present, and 0: no tie present
Binary ties make up the 2 mode (actor x institution) asymmetric network matrix
2 mode data transformed into a 1 mode (actor x actor) symmetric network
matrix to run analysis
Ties are undirected
5. Research Question 1: Methodology
Networks Analysed:
Total network: All actors as nodes, with ties (both top-down and bottom-up),
in a transformed 2 mode to 1 mode network
Sub-networks: Either top-down or bottom-up institutional designs, are also
made up of actors as nodes, with ties, in a transformed 2 mode to 1 mode
network
> Method- Research Question 1
Social Network Analysis (SNA): three measures of centrality to assess
embededness:
Degree centrality: the number of ties an actor directly shares with others in
the network
Betweeness centrality: the number of times an actor connects two
disconnected actors
Eigenvector centrality: higher when an actor is connected to institutions that
are also well connected
6. Research Question 1: Methodology
> Analysis:
T-test: compare the mean centrality: degree, betweeness and eigenvector of
national government actors and all other groups (total network and top-down
designs), and local government actors and all other groups (bottom-up
designs)
Two-way ANOVA: tests the level of significance of differences in normed
centrality: degree, betweeness, and eigenvector means between the 7 actor
types, with the p-value
Regression model: to fit the data and estimate the significance and strength
of the relationship between independent variable actor type and dependent
variable actors’ inclusion (dependent variable: centrality)
If an actor has a high degree centrality, betweeness centrality and eigenvector
centrality, relative to others = actor is well embedded in the network
7. Table 3. Normed betweeness centrality means for the total network and sub-networks
8. Research Question 1: Results
> Hypothesis: National actors are well embedded in top-down institutional designs,
and local actors are well embedded in bottom-up institutional designs
National governments do not have a high degree centrality (Table 2- Annex),
but high betweeness centrality (Table 3) and eigenvector centrality (Table 4-
Annex) in top-down designs
Local actors have a slightly higher degree centrality, no betweeness centrality,
but a high eigenvector centrality, in bottom-up institutional designs
No significant effect of actor type on degree, betweeness, eigenvector
centrality
Conclude: National actors are relatively better embedded in top-down
institutional designs, and local actors are relatively better embedded in
bottom-up institutional designs
9. The Relevance of Clustering
Research Question 2: Part A and B
> Clusters of actors collaborate inside closed groups, having repeated transactions
leading to trust, thus supporting cooperation
> Core/Periphery- Actors in the core are more able to coordinate their activities
> Aim: Examine whether clustering occurs according to actor type in climate change
adaptation institutions in a complex network in Bhutan, by applying a multilevel
governance framework to explore links between actors and institutions
> Research Question 2. Part A: Do actors of the same type cluster together within
networks?
Hypothesis: Actors of the same type collaborate by clustering; the density
between within-group ties and outside-group ties are significantly different
Independent Variable: Actor type
Dependent Variable: Clustering
> Research Question 2. Part B: Which actors make up the core and periphery of
the network?
10. Research Question 2: Part A and B
Methodology
> PART A & B: An MDS plot is used to initially locate clusters:
Actors located in the centre of the cluster = core actors
Actors located further away from the centre = peripheral actors
Analysis: Total network (both top-down and bottom-up institutional designs)
> Part A: Clustering analysed via tie density:
Structural Block model option of an ANOVA Density model: test whether within
and between group ties are significantly different across actor types
> Part B: Location via a simple Core/Periphery model:
Core actors = highest density of ties amongst themselves, collaborating in
common institutions
Peripheral actors = lower density of ties amongst themselves, fewer
institutions in common
11. Figure 5. An MDS plot shows the total network.
Red circles: international organisations, orange: foreign government, yellow: national
government, green: local government, blue: NGO, purple: corporation, pink: community
12. Table 5. Structural Block model option of an ANOVA Density model (a) with model fit (b), and a
Simple Core/Periphery model with density matrix table (c) for the total network
13. Research Question 2: Results
> Part A: Do actors of the same type cluster together within networks?
Hypothesis: Actors of the same type collaborate by clustering; the density
between within-group ties and outside-group ties are significantly different
Results: There is a statistically significant effect of variable “actor type” on
variable “clustering” in the network; accept that the density between within-
group ties and outside-group ties are significantly different
> Part B: Which actors make up the core and periphery of the network?
Results: Many national level actors in the core
> Conclude:
Actors of the same type collaborate by clustering
National actors cluster and occupy the core of the network, sharing many ties
and cooperating with other actor types
14. The Relevance of Fragmentation
Research Question 3
> We define fragmentation where there is a high proportion of actors unable to reach
each-other in the network:
Vertical fragmentation: Between institutions at different scales such as from
national to local
Horizontal fragmentation: Between institutions in different sectors such as
forestry or water
> Fragmentation in institutions:
Could affect actors’ ability to implement projects effectively, through inhibiting
joint decision-making
Creates barriers to cooperation
> Aim: Examine fragmentation in climate change adaptation institutions in a complex
network in Bhutan, by applying a multilevel governance framework to explore links
between actors and institutions within networks
> Research Question: Does the network display fragmentation?
15. Research Question 3 - Hypothesis and
Methodology
> Hypothesis: Top-down institutional designs are less fragmented than bottom-up
institutional designs
> Fragmentation analysis:
Networks: the total network (both top-down and bottom-up institutional
designs), and sub-networks; top-down and bottom-up
Fragmentation score: the proportion of actors unable to reach each-other
Compare the fragmentation score of institutional design relative to each-other
> Density score:
Equal to 1 = all actors within the network are tied directly to each other
Equal to 0 = network is fully disconnected and therefore fragmented
16. Research Question 3: Results
Table 6. Fragmentation score for the total network, and sub-networks; top-down or bottom-up
institutional designs
> Results:
Low amount of fragmentation in top-down designs
High amount of fragmentation in bottom-up designs
Overall density scores (Table 1- Annex) are low for each institutional design
(top-down = 0.035, bottom-up = 0.052)
Conclude: Few ties through which information can flow, both networks are
disconnected and therefore fragmented
17. Conclusion
Some difference between top-down and bottom-up institutional designs:
National actors relatively better embedded in top-down, local actors relatively
better embedded in bottom-up (Table 2- Annex, Table 3, Table 4- Annex)
Lack of inclusion of local actors (Table 2, 3, 4)
Some actors may work on climate change projects in isolation- UNFAO (Fig.
5)
Actors of the same type collaborate by clustering (Figure 5) (Table 5)
National actors cluster/core of the network, sharing many ties/cooperate with
different actor types (Table 5)
Top-down institutional designs less fragmented than bottom-up, but with
overall low density scores for each (Table 1-Annex). Conclude that both
demonstrate fragmentation (Table 1- Annex) (Table 6)
The total network in Bhutan is fragmented (Table 1- Annex) (Table 6), but
displays clustering (Table 5); conclude that information sharing between actors
to plan, monitor and enforce climate change adaptation projects may be low
18. Recommendations
> General recommendations:
Increased involvement from national government actors: more influence, ability
to collaborate and capacity to coordinate in the network
> Recommendations for a “polycentric” approach to governance:
A hybrid approach between top-down, and bottom-up designs, with national
government providing a guiding framework, allowing local communities to
make implementation decisions based on community knowledge
Increase both horizontal and especially vertical collaboration, to increase local
actor inclusion
Support vertical and horizontal integration, to decrease fragmentation