Version classiqueVersion mobile

The role of German universities in a system of joint knowledge generation and innovation

 | 
Mirja Meyborg

5. The Concept of Social Network Analysis

Texte intégral

  • 26 The term actor can imply countries, regions, institutions or individuals and is used interchangeabl (...)

1The SNA is a useful instrument for measuring the nature and extent of collaboration activity among economic development practitioners within a metropolitan area (Reid and Smith 2009). According to Tindall and Wellman (2001), a SNA implies the study of social structure and its effects, whereas the social structure is regarded as a social network which consists of a set of actors26 and a set of relationships that connect the actors. Hence, the SNA concentrates on relationships between actors and explores the availability of resources and the exchange of resources between these actors. The resources can be of two different types, namely tangible or intangible resources. Tangible resources are, for example, goods, services, or money, whereas intangible resources consist of information, social support, or influence (Scott 1991). As this PhD thesis is about knowledge generation and innovation, it concentrates on intangible resources.

  • 27 Forfurther information see http://www.analytictech.com/ucinet/, 21.10.2010.
  • 28 Forfurther information see http://pajek.imfm.si/doku.php?id=start, 23.05.2008.
  • 29 For further information see http://www.analytictech.com/netdraw/netdraw.htm, 10.05.2011.

2However, by means of SNA, it can generally be illustrated how countries, regions, institutions or individuals cooperate with each other, thus, demonstrating the development of the regional, national or international connectivity. So, the SNA method is designed to “[...] discover patterns of interaction between social actors in social networks” (Xu and Chen 2005). It implements this by revealing the overall network structure, as well as that of subgroups within the network, then examining the patterns of interaction among these various groups. It is an interdisciplinary methodology developed mainly by sociologists and researchers in social psychology. Later on, the SNA has been further extended in collaboration with mathematics, statistics, and computing. Especially advances in computer technology, availability of computer databases and the emergence of several software packages like Ucinet27, Pajek28 or Net Draw29 has allowed for the construction and analysis of scientific collaboration networks. All this made it attractive also to other disciplines like economics or industrial engineering as it now developed to a formal analyzing tool (Cantner and Graf 2006). Hence, the SNA has quickly developed over the last two decades as the fruitful combination of theoretical concepts with the numerous application possibilities has attracted many other research disciplines (Wassermann and Faust 1994).

  • 30 See Wassermann and Faust (1994): Social Network Analysis; Cambridge University Press for further in (...)

3The networks illustrated by the SNA are mostly described by agraph30, consisting of nodes joined by lines, ie. the SNA treats individuals as nodes and the relationship between individuals as linkages (Wasserman and Faust 1994). Thereby, graph theory does not consider the conventional representation of physical distance, but measures the distance between nodes solely in terms of the number of lines which it is necessary to traverse in order to get from one node to another (Scott 1988). In the following, it is illustrated how a network can be constituted:

Image 10000000000003F8000001DB18C51F33.jpg

Figure 6: Star- and Line-Network (Nooy et al. 2005)

4Figure 6 shows two possible forms of a network. On the left side, a so-called star network is pictured. Such a network is characterised by only one central node (node 5) which is linked to all other nodes (nodes 1 to 4). Thus, node 5 possesses four distinct linkages to other actors, whereas nodes 1, 2, 3 and 4 have just one distinct linkage to node 5. In this context, it is feasible to determine which node the most centralised one is, namely node 5. On the right side, a so-called line-graph is illustrated. In this case, nodes 8, 9 and 10 have two distinct linkages each, whereas nodes 6 and 7 possess only one linkage each. Hence, it is now much more difficult to find out which node is t h e most centralised one. Therefore, network characteristics have to be used in order to find out which actor might be the most linked and centralised one within a network (Nooy et al., 2005). Thus, centrality measures help to explore, if an actor who is highly centralised is also responsible for the support of innovations. Thereby, centrality can be measured by means of different methods. There are basically three applied types of centrality, namely degree centrality, betweenness centrality, and closeness centrality.

5On the one side, the degree centrality of an actor is important as this measure implies the simplest possibility to determine the actors' centrality. Here, it is defined that central actors must be the most active in the sense that they have the most linkages to other actors in the network. Degree centrality is expressed as follows:

6CD (ni) = d (ni)/(g-1)

7CD ( ni) -Degree-centrality

8d (ni) - Number of linkages (degree)

9g - Number of actors of t h e network (size of the network).

10The higher the degree-centrality the more active is the actor within the network and can use its position to influence others agents or to get more information. Degree centrality can be seen as a measurement of communication activity of an agent. The following network shows the actor with the highest value of degree centrality:

Image 10000000000004E00000023667AB4634.jpg

Figure 7: A social Network - highest Degree Centrality (own illustration).

  • 31 Connected graphs are graphs without isolated nodes.

Second, the centrality of an actor is based on closeness or distance, focusing on how close an actor is to all other actors in the set of actors. Thus, if an actor can quickly interact with all other actors, it possesses a high value of closeness centrality and is not reliant on any other actor for the relaying of any information. The concept of closeness centrality is available only for strongly connected graphs31. The value of closeness centrality is calculated as follows:
Image 10000000000001F9000001102358367A.jpg

11where actor closeness centrality is the inverse of the sum of geodesic distances from actor / to the g-1 other actors, and can be normalized by g-1 (Wassermann and Faust, 1994). The higher the value of closeness centrality the less an actor has to rely on the transmission of any information from any other actor. The highest value of closeness centrality within a network can be pictured as follows:

Image 10000000000004C1000002401B8561D6.jpg

Figure 8: A social network - highest Closeness Centrality (own illustration).

  • 32 The difference is that degree centrality only considers direct relationships and closeness centrali (...)

While degree and closeness centrality always regard double relationships of the considered actor32, the value of betweeness centrality considers three actors. It measures the extent to which an actor is needed as a link in the chains of contacts, facilitating the spread of information through the network. It counts how often one actor lies on the shortest path between two other actors, hence, taking into account the connectivity of the node's neighbours, giving a higher value for nodes which bridge clusters (see Nooy et al., 2005). The value of betweenness centrality is defined as follows:
Image 10000000000002060000009EF8E2D8E0.jpg

gjk (ni) -Number of shortest path between two nodes j and k (geodesies) where node n, is involved.
gjk - Number of shortest paths between two nodes j and k
Cb (ni) - Betweenness centrality

12The higher the value of betweenness centrality the more important is the actor within the network, i.e. the actor possesses a strong transmitter function within the network so that he is strongly needed for the transmission of information. The following figure shows two actors within a network which possess both the highest value of betweenness centrality:

Image 1000000000000518000001E51A60DB9E.jpg

Figure 9: A social network - highest Betweenness Centrality (own illustration).

13Again, as it is intended to demonstrate how strong all German universities are generally linked to other actors within the knowledge and innovation networks, the concepts of degree and betweenness centrality are used for further analyses within this PhD thesis. As already mentioned above, the emergence of several software packages like Ucinet, Pajek or NetDraw has allowed for the construction and analysis of scientific collaboration networks which contain huge mounds of data. Here, the software package Pajek is used in order to draw networks and to calculate above mentioned centrality measures. However, before doing this, the publication and patent data had to be prepared in a special manner. First of all, they had to be set up as follows:

Image 100000000000057F00000194934969A4.jpg

Table 5: Example of Patent Data Set-up for Data Preparation with Java (own illustration).

Image 10000000000005680000019810A6B7B7.jpg

  • 33 As can be seen from the table, the columns application sequence number and inventor sequence number (...)

Table 6: Example of Publication Data Set-Up for Data Preparation with Java (own illustration).33

  • 34 The chosen example is according to the example illustrated in the dissertation of Bertram (2011).

14A Java-written program further prepared both data sets in such a manner that they could be used for drawing networks and calculating degree and betweenness centrality. Thereby, the linkages of each actor are weighted, ie. each linkage of an actor to a patent or publication is divided by the sum of all linkages that appear on the patent or publication. In doing so, the sum of all linkages for one patent or publication is always 1. The final value of an actor then results from the sum of all its weighted linkages that appears on any patent or publication. The following example clarifies this approach by means of publications.34

Image 10000000000004F4000000BBD949DEAB.jpg

Table 7: An Example of a Social Network - Part 1 (own illustration).

15The network of this example consists of four actors (A, B, C, D) who have the following weighted linkages:

Image 10000000000002EC000000D284F9DFB2.jpg

Table 8: An Example of a Social Network - Part 2 (own illustration).

16The network of this example looks as follows and shows all corresponding linkages of actor A, B, C and D, whereas actor A and actor D have no linkage to each other.

Image 100000000000027D00000249CCF65E70.jpg

Figure 10: An Example of a social Network - Part 3 (own illustration).

17After having presented how the networks of publications and patents are constituted, the following chapter now introduces the concept of the spatial analysis that has been applied in the course of this PhD thesis.

Notes

26 The term actor can imply countries, regions, institutions or individuals and is used interchangeable in the course of this PhD thesis.

27 Forfurther information see http://www.analytictech.com/ucinet/, 21.10.2010.

28 Forfurther information see http://pajek.imfm.si/doku.php?id=start, 23.05.2008.

29 For further information see http://www.analytictech.com/netdraw/netdraw.htm, 10.05.2011.

30 See Wassermann and Faust (1994): Social Network Analysis; Cambridge University Press for further information on graph theory.

31 Connected graphs are graphs without isolated nodes.

32 The difference is that degree centrality only considers direct relationships and closeness centrality also indirect linkages.

33 As can be seen from the table, the columns application sequence number and inventor sequence number are only needed for the patent data, as in this context, it is possible to explore co-applicant or co-inventor networks. For the preparation of the publication data, both columns are not required as it is always about co-authorship, no other classification could be made in this regard.

34 The chosen example is according to the example illustrated in the dissertation of Bertram (2011).

CC-BY-SA-4.0

Le texte seul est utilisable sous licence CC BY-SA 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.

Lire

Open access

Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search