7. Homophily in stable partnerships
One of the goals of Horizon 2020, as of all European funded research, is to bring about robust collaboration between universities (seen as the natural locus for technological research) and private companies (seen as the natural locus for technological development of that research). This is clearly happening, in some cases. In what follows we try to quantify the extent to which it is happening.
We take the stable partnership graph as the network of all non-casual partnerships in Horizon 2020. It has 46,362 edges, each representing one such partnership in one direction (recall that, because of computational requirements, the graph stores two edges for each partnership between A and B, (A => B) and (B => A). Of these, 21,366 (46%) connect partners of the same type.
In the following table, we show the percentage of partnership with orgs of the same type in all different types.
| Type | % edges w/orgs of same type (actual) | % edges w/orgs of same type (random null) | % Difference |
|---|---|---|---|
| PRC | 45 | 40 | +5 |
| HES | 59 | 18 | +41 |
| REC | 38 | 22 | +16 |
| PUB | 46 | 10 | +36 |
| OTH | 14 | 8 | +6 |
| ALL | 46 | 26 | +20 |
The third column shows the percentage of partners of the same type that each type of organisations would have if it drew partners at random; the fourth column shows the difference in percentage between the observed choice and random choice (which we use as a benchmark). Companies and the residual category OTH do not appear to have a strong preference for choosing partners of their own respective kinds, whereas the other three categories do.
The stable partnership graphs has 46,362 directed edges. Each partnership gives rise to two directed edges: (A => B) and (B => A). These can be interpreted as decisions to make a partnership. Of these, only 3,254 (7%) involve one university (activityType = HES) and one company (activityType = PRC). If organisations picked partners at random, the share of partnerships involving one university and one company would be exactly double: 14%. This means that their active choice of partners lead them away from the type of partnership most desirable by the standards of policy makers.
In the graphs at the end of this page, the edges of this type are coloured red. It is visually apparent that stable partnership between universities and companies are relatively few in Horizon 2020, though of course it is difficult to say what a "right" or "desirable" proportion is.
We conclude that the stable partnership network in Horizon 2020 shows marked homophily (also known as assortativity) on activityType. Furthermore, different types of organisations have different behaviours in this respect. Universities are the most strongly assortative organisation type: they display a marked preference for building stable partnerships with other universities.
