Single Review on past events

What do pensions, Stuttgart 21 and software patents have in common? – There are policy debates about them amongst followers and opponents, and Philip Leifeld develops methods to analyze them.

“Political discourse takes place when political actors, like interest groups, politicians, parties etc., make public statements about their preferred policies conditional on each other in order to affect the policy outcome”, Philip Leifeld explained. “Actors take part in a political debate to reduce their own uncertainty or to convince other actors.”

In order to analyze these discourses, Philip Leifeld wrote a software called “Discourse Network Analyzer” (DNA), which can be used to annotate newspaper articles or other text data by encoding the opinions and statements that political actors express. Based on these extracted statements, the software can build networks of actors and political concepts and finally serve to identify clusters, or “discourse coalitions”, of actors in a debate. In these discourse networks, the similarity of actors' positions is determined by their agreement on, or joint rejection of, political concepts.

Since actors with official institutional positions (governmental actors) make more statements in the media than others and thus occupy more central positions in the discourse network, the software can manage to normalize the similarities and thus eliminate this potential source of bias. The final discourse network can be visualized by imposing a threshold value on the similarities of actors.

As one example, the Postdoctoral Fellow presented his research on German Pension Politics (“Riester-Reform”) up to the year 2001 based on articles in the Frankfurter Allgemeine Zeitung (FAZ). He manually encoded 7.249 statements of 246 organizations and/or 461 persons in 1.879 articles between 1993 and 2001. The statements of these actors were classified into 68 concepts. Using his DNA approach, Philip Leifeld could demonstrate how a dominant discourse coalition in the early 1990s was gradually replaced by a challenging coalition composed of banks, insurance companies and employers' associations with different ideological claims before the reform was adopted in 2001.

Another example of Discourse Network Analysis was Philip Leifeld's work on the European conflict on software patents (joint work with Sebastian Haunss). Software patents were proposed by the European Commission and large companies in 1997, but the initiative was rejected by the European Parliament in 2005. In the meantime, the issue was debated in the media. Philip Leifeld wanted to find out how a coalition of small- and medium-sized enterprises and social movements could prevail over a strong industry-led coalition of supporters of software patents (SWP).

He measured the evolution of discourse coalitions as portrayed in five major EU newspapers between 1997 and 2005. His analysis of 33 categories/arguments/claims (“software patents are good/bad because…”), distributed over 450 statements, could provide evidence that the anti-SWP coalition had a larger constituency and was structurally more congruent at all times. The same thing could be shown for the concept networks which might be an explanation for the success of the SWP opponents.

With the implementation of a new statistical model for network data – relational event models for signed bipartite graphs –, Philip Leifeld is planning to answer interesting questions beyond the descriptive analysis of case studies, such as: Is there reciprocity among actors when they contribute statements? Do they punish each other via statements in the discourse? Is there social balance among actors? The development of these methods is supported by a new PhD project on water policy debates under his supervision at the Swiss Federal Institute of Aquatic Science and Technology (Eawag) in Zurich.