Project C4:

Measuring a common space and the dynamics of reform positions

 

Coordinators:

Prof. Nicole Baerg, Ph.D. / Prof. Thomas Gschwend, Ph.D. / Prof. Dr. Simone Paolo Ponzetto / Prof. Dr. Heiner Stuckenschmidt

 
 Researchers:

Jason Eichorst, Ph.D. / Sanja Stajner, Ph.D.

 Associates:
James Lo, Ph.D. / Prof. Sven-Oliver Proksch, Ph.D. / Will Lowe, Ph.D.


 Duration:01.01.2010 - 31.12.2017

 

Summary

The C4 project is tasked with developing useful quantitative tools that make spatial placement of political actors possible. Having comparative measures enables researchers to test theories and has proven powerful in explaining important attributes of the political economy of reforms.


One typical problem in empirical political science research is estimating ideal points, policy preferences, topics, or issue emphases across countries on a common scale. C4 project members are concerned with modeling such things. For example, there are two sources of bias-in survey research that generates obstacles for making cross-national comparisons of voters and parties on a common scale. Research by C4 project members presents a scaling solution that enables the estimation of voters, national parties, and political groups on a common left-right scale. Applying this method to surveys in United Kingdom and European Parliament in 2009, this research shows high correlation with expert surveys, but a completely different configuration of political parties when compared to uncorrected scaling methods. Adequately capturing the configuration of political parties is critical for generating a reliable measure. The results from this paper are presented in the figure below.


Another typical problem is how to incorporate textual information as data. Political speech, for example, is useful for mining information about elites and voter preferences and opinions. Work within the C4 project brings together topic models and scaling models to make this possible. One feature of C4 research that communicates to this objective is developing theoretical models that structure the underlying data generating process of speech. Work by Lowe & Benoit (2013) demonstrate the usefulness of modeling political text to uncover scaled preferences by matching human approaches to machine approaches that both estimate elite issue positions from the 2009 Irish budget parliamentary debates. The figure below shows similarities between the two different approaches. This research reveals that text models can help researchers efficiently obtain reliable measures of ideological preferences from partisan speeches.

The C4 research team shares the common objective: to strengthen empirical tools that spatially capture the placement of actors in a common space. Using mass surveys, political speeches, and roll call data, we apply our empirical tools to test theories in varied substantive fields, including political economy, political representation, partisan organizations, and mass voting behavior. Click here to view a list of our relevant publications.