My research agenda focuses on political networks, particularly interest groups and state actors. Policy networks are an important and fascinating research topic because they influence political outcomes and are yet poorly understood due to the complex interdependencies between units of observation. Policy networks often act in the shadow of existing hierarchies and government institutions, and uncovering and explaining these structures and their impact is key to understanding how political outcomes come about. Think, for example, about the factors that determine lobbying success of interest groups in national policy making, e.g., in various environmental or social policy domains. Or consider the emergence of organizational preferences about what should be done to solve complex problems like non-point source pollution or population aging. Chances are that dependencies between political actors crucially matter for our understanding of these phenomena.
At the same time, the dependence between observations inherent in network data poses severe statistical challenges that need to be solved for proper theory testing. Therefore my research combines substantive theory development and testing with the development of new computational and statistical methodologies suitable for network data.
My publications and projects cover topics such as information and resource exchange between local or national-level political actors in decision making and implementation, the endogenous formation of policy instrument preferences and policy beliefs, the structure of environmental policy subsystems, collaborative governance and the role of policy forums, polarization and conflict in legislative and decision-making bodies and in transnational and international environmental politics, and the micro-level processes that govern policy debates. I study these topics in empirical fields, such as climate, water, toxic chemicals, social and pension politics, and international cooperation and conflict, but also from a more general methodological perspective or governance perspective.
Organized Interests and Policy Networks
The influence exerted by organized interests through networks is often more subtle than is commonly assumed. In my first published study with the title "Information exchange in policy networks" (published jointly with Volker Schneider in the American Journal of Political Science), I analyzed the communication patterns between interest groups and state actors in German toxic chemicals regulation and developed and tested an explanation based on transaction cost theory. I could demonstrate empirically that joint memberships of political actors in institutionalized policy forums beyond the control of the state facilitate tie formation. Actors pursue a variety of strategies to establish potentially valuable contacts without incurring too many costs, therefore policy forums are one out of several cheap ways to get in touch.
In a follow-up paper with the title "Policy forums: why do they exist and what are they used for?" (published jointly with Manuel Fischer in Policy Sciences), I subsequently developed a theory of the existence, size, and shape of such institutionalized forums, drawing on the idea of asymmetric multipartite exchange.
In an article with the title "Structural and institutional determinants of influence reputation: A Comparison of Collaborative and Adversarial Policy Networks in Decision Making and Implementation" (published jointly with Karin Ingold in the Journal of Public Administration Research and Theory), we show that joint forum memberships are also an important resource for actors with regard to reputation formation—in collaborative governance and adversarial settings, in decision-making and implementation networks, and at the local as well as the national level. Joint memberships of two actors cause increased perceptions of mutual relevance and boost the overall reputation of actors through network diffusion. Similarly, occupying structurally salient positions in the collaboration network increases actors' reputation, and influence reputation has been shown to translate into actual policy-related influence, which makes these factors highly consequential for political outcomes.
In a working paper with the title "Contributions by Interest Groups to Lobbying Coalitions" (joint work with Michael T. Heaney), we consider not only memberships in forums, but also whether and why organizations who are members of lobbying coalitions provide leadership contributions to their respective coalitions. This is driven both by factors pertaining to the composition of the coalition and the individual interest group. Moreover, we find network effects through which leadership in one coalition affects leadership in another coalition.
Transnational and International Networks
Similar processes can be at work in policy-revelant international or transnational expert committees: In a paper in progress on "Self-Reinforcing Recruitment Processes in an Epistemic Community" (joint work with Dana R. Fisher and Joe Waggle), I analyze the nomination and recruitment patterns within a large epistemic community founded by the United Nations. Like in national policy forums, network effects have a potentially distorting effect on policy (advice): A small group of well-connected transnational elites has enough agenda setting power to determine the composition of the whole group of hundreds of experts through a biased snowball nomination system.
I am also interested in network effects in more formal international and supranational governance arrangements: In another paper, which won a best conference paper award, I examine national parliamentary coordination in the "early warning system," a large-scale governance arrangement introduced in the European Union with the Treaty of Lisbon in 2009. The main finding is that social influence (rather than selection) between parliamentary chambers along the lines of party family majorities and ideology determine whether multiple national parliaments co-veto legislative proposals by the European Commission. For this paper, we implemented relational event models for bipartite networks and came up with a permutation approach to disentangle social influence from social selection. This work is co-authored by Thomas Malang and my PhD student Laurence Brandenberger.
At the international level more generally, I have co-authored an article with the title "Common allies and their role in suppressing bilateral conflict" (under review; joint work with Aisha Bradshaw, Weihua Lia, Caitlin Clary, and Skyler J. Cranmer), where we examine the suppressing effect of states' embeddedness in shared and multilateral alliances on militarized interstate disputes. While my research profile focuses mostly on national political processes and organized interests, there are interesting theoretical similarities because variables like intermediary organizations and leadership play important and partly similar roles at both levels.
Discourse Network Analysis
One of my longstanding research interests is the analysis of policy debates on topics such as climate change or a sustainable design of the pension system. State actors and private actors speak about their policy instrument preferences in the media, parliamentary hearings, or other public venues, and react to each other's statements in complex ways, resulting in coalitions and other interesting configurations at the macro level. Disentangling these partly endogenous processes using advanced computational and statistical approaches permits me not only to understand how policy debates work and how topics get on or off the agenda, but ultimately also what laws and regulations are adopted and how an issue is framed by relevant decision-makers. A primer on discourse network analysis is forthcoming in the Oxford Handbook of Political Networks. I have recently published a monograph on discourse networks and their application to German pension politics with Campus (distributed internationally through the University of Chicago Press). The title of the book is "Policy Debates as Dynamic Networks: German Pension Politics and Privatization Discourse". The book is based on my dissertation and covers discourse network analysis in detail. You should get a copy.
I have applied my discourse network methodology and my software Discourse Network Analyzer (DNA) to a range of empirical policy debates, including pension politics (see also my paper in the Policy Studies Journal), software patents (see my paper with Sebastian Haunss in the European Journal of Political Research), and climate change (see my joint work with Dana R. Fisher and Joe Waggle in the American Behavioral Scientist and in Climatic Change).
Together with my PhD student Laurence M. Brandenberger, I am now working on theoretically informed statistical models to understand the micro processes guiding policy debates and in order to forecast how policy debates develop in the near future (given the observed temporal network sequence of statements by political actors up to the present day). This will be tremendously helpful in understanding and predicting public opinion changes and elite discourse on salient topics such as climate change, immigration, or the design of a sustainable pension system.
A first step in this direction has been the publication of an agent-based simulation model on the evolution of policy debates, which uses the discourse network analysis framework for model validation. The paper with the title "Polarization of Coalitions in an Agent-Based Model of Political Discourse" was published in the open access journal Computation Social Networks. It shows how some simple micro-level processes in a two-mode network of actors and concepts lead to aggregate patterns like discourse coalitions or advocacy coalitions and polarization. The paper has been one of the journal's most widely read articles since the journal's inception in 2013.
Inferential Network Analysis
The analysis of dependencies in political networks poses interesting methodological challenges because standard regression models (e.g., the generalized linear model or mixed-effect models) as well as dyadic approaches to data analysis (like time-series-cross-sectional models) would yield biased estimates and uncertainty measures if applied to network data. The problem is ubiquitous because dependent data can be found in nearly all domains of political research, whether they carry the label "network" or not. Moreover, these models do not permit the user to test relational theories, which are often required to understand political phenomena. The current limits of statistical methodology inhibit proper relational theory building. However, I believe embracing a relational view is important as political science is inherently relational. Therefore I actively develop and implement inferential models for dependent data in order to support my substantive and theoretically driven research.
More specifically, I have been working on extensions and implementations of the temporal exponential random graph model (TERGM), the temporal network autocorrelation model (TNAM), the relational event model (REM) for bipartite signed graphs, and related approaches. I have developed or extended these models, implemented them in my own R package called xergm and several sub-packages, and I conduct Monte Carlo simulation studies in order to identify the precise conditions under which they offer advantages over existing models. Several papers about these developments are currently under review in various methodology journals. The first paper from this series of papers that has been accepted for publication has the title "Navigating the Range of Statistical Models for Inferential Network Analysis" (joint work with Skyler J. Cranmer, Scott D. McClurg and Meredith Rolfe) and is forthcoming in the American Journal of Political Science.
Besides xergm, I am the author of another widely used R package called texreg. The package serves to convert statistical model output (also of network models) to LaTeX, HTML, MS Word, ASCII text, or Markdown tables, with multiple models arranged side-by-side. I have published an open-access article with the title "texreg: Conversion of Statistical Model Output in R to LaTeX and HTML Tables" in the Journal of Statistical Software.