Intermediation and network constraint

The stable partnership graph allows us to detect organisations potentially vulnerable to being intermediated. These are organisation that are part of the giant component of the EU funded research network, but through only one stable partner. “Repeat business” with only one partner is what we would expect to see from organisations that are dependent on that one partner to win grants (though, of course, there are other possible interpretations for the same pattern). This line of inquiry is grounded in Burt’s (2003) theory of network constraints. Constrained agents in social networks have limited options, because their access to the broader network has to go through a single agent. That agent might be tempted to exploit their position to extract rents from it.

In what follows, we call “intermediated” an organisation that (a) is part of the giant component of the stable partnership network of a framework programme; (b) is only connected to one other organisation in it. We call “intermediary” an organisation that is the only stable partner of at least one intermediated. Some intermediaries intermediate more than one organisation.

Detecting intermediated nodes is a harder problem that it seems at a first glance. Obviously all organisations of degree 1 in the stable partnership graph are intermediated, because they their only pathway into the network is through their only neighbour. Determining the intermediated status of nodes with degree n>1 is a recursive problem, computationally hard. We left it for another project. We identified intermediated nodes as follows. First, mark as intermediated all nodes with degree 1; second, visually inspect the graph for peripheral structures that access the broader network through one single chokepoint. “Centerward” neighbors of intermediated nodes were classed as intermediaries: all non-intermediate, non-internediary organisations were classed as unconstrained.

In the table below, “max intermediated” is shorthand for the maximum number of organisations intermediated by a single intermediary.

	intermediated	intermediaries	max intermediated
FP7	1,535 (17.5%)	729 (7.9%)	32 (Fraunhofer Gesellschaft)
H2020	470 (21.7%)	240 (11.3%)	48 (Fraunhofer Gesellschaft)

Intermediation and topology

Here is a picture that shows the nodes in the giant component of the H2020 graph color-coded by their role in with respect to network constraints. Bright green nodes are intermediated; red ones are intermediaries; dark green ones are unconstrained, neither intermediated nor themselves intermediaries.

Unsurprisingly, intermediated orgs are largely disconnected from each other. Both unconstrained orgs and (even more so) intermediaries are, however, quite densely connected. The intermediation structures of FP7 and H2020 appear to be similar. The organisation functioning as the single point of access for most intermediated is Fraunhofer Gesellschaft in both framework programmes. It intermediates 32 in FP7 and 48 in H2020.

subgraph	nodes	edges	% nodes	% edges
H2020: intermediated	470	23	22.7%	0.1%
H2020: unconstrained	1,359	5,514	65.7%	35.0%
H2020: intermediaries	240	2,599	11.6%	16.5%
FP7: intermediated	1,535	3	17.5%	0%
FP7: unconstrained	6,474	29,720	74.1%	30.8%
FP7: intermediaries	729	16,996	8.3%	17.6%

The percentages of edges do not add to 100% because they only account for intra-subgraph connections. Edges between organisations of two different types (for example intermediated and unconstrained) are dropped from the count.