Using network analysis to understand what makes an organisation

Earlier in the year, Dr Johan Koskinen gave a talk on his Bayesian modelling that highlights how imperfect information can be used to construct networks of organisation. In particular, he focused on how information about the inter-organisational ties can be altered, depending on the how the information is acquired.

When attempting to understand an organisational network, it is tempting to rely on organisational representatives as informants for constructing inter-organisational ties. Much of the research relies on these reports and it is rare to find research on independently defined organisational ties, such as financial transactions or collaborative agreements. This is a problem, as it is not clear how individual reports relate to actual organisational behaviour.

In this talk, Dr Koskinen presented a measurement model that assumes the true inter-organisational network is an unobserved and undirected network, but that error lies in the reports of the network by the organisational representatives. This model draws on the cognitive social structures data collection paradigm and test-theory without an answer-key, and the organisational network is assumed to follow an exponential random graph model a priori. The target of inference is two-pronged: simultaneously aiming to infer the network mechanisms of tie-formation of the unobserved network and to assess the competencies and cognitive biases of the informants.

The resulting methodological framework can be seen as explicitly modelling the famous duality of individuals and groups. Dr Koskinen illustrated the procedure using data on organisations involved in transportation and cross-border collaboration in a number of European countries, where the number of organisational informants range from single individuals to a few organisational members.

We thank Dr Koskinen for sharing his expertise on organisational networks.